A method for analyzing and improving performance of unit AGC and line CPS index correlation

Abstract

The application relates to a method for analyzing and improving the performance of a unit AGC and a line CPS index. The correlation between the unit automatic generation control response characteristic and the line CPS index is established, on the basis of which, key units are identified, the unit performance is sorted, the AGC performance of the units is improved in a targeted manner, and the CPS level of the tie line is further improved.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method for correlation analysis and performance improvement of unit AGC performance and line CPS index. Background Technology

[0002] Automatic Generation Control (AGC) systems play a crucial role in tie-line power regulation. Currently, a certain province is at a relatively high level in industry benchmarking for regional power grid control performance standards (CPS). With the gradual advancement of large-scale offshore wind power development and construction in this province, the proportion of its renewable energy installed capacity is increasing year by year. In the foreseeable future, the system will mainly face the following challenges:

[0003] First, balancing supply and demand is becoming increasingly difficult. New energy power generation exhibits significant meteorological characteristics, including randomness, volatility, and intermittency. It also suffers from the reverse peak-shaving supply characteristics of "extreme heat with no wind" and "no sunlight during the evening peak," and the capacity for transmitting and allocating surplus power is insufficient. For example, from September 28th to October 2nd, 2021, a certain province had 5,893,250 kW of wind power installed capacity, with an average output of 345,000 kW (6.0%) and a maximum output of 1,631,000 kW (28.3%). On October 1st, the province's wind power output ranged from 300 kW to 336,000 kW (0%-5.8%). Meteorological conditions exacerbated the randomness and volatility of new energy power output, placing higher demands on the system's regulation capabilities.

[0004] Secondly, the installed capacity and proportion of new energy sources are increasing year by year, which places higher demands on the flexible operation of traditional power sources. By 2030, the installed capacity of wind power in the province is expected to exceed 15 million kilowatts, accounting for nearly 20%. On cloudy days with no wind, the power shortage will reach tens of millions of kilowatts. As the "ballast" of the power system, thermal power and hydropower units will play a more important regulatory role. Fully tapping the active support capacity of the units and improving their flexible operation capabilities are important issues that power system operators urgently need to address.

[0005] Third, inter-provincial interconnected power grids offer greater flexibility in power transmission. Taking a certain regional power grid as an example, the installed capacity of new energy sources in various provinces continues to rise, and the significant differences in inter-provincial meteorological conditions lead to prominent power exchange phenomena in inter-provincial interconnection lines, placing higher demands on the control capabilities of dispatching and operation personnel. Currently, with the large-scale grid connection of new energy units, their installed capacity has reached 14% of the province's total power system installed capacity, while pumped storage and peak-shaving gas-fired power generation only account for 8%. The current installed capacity of new energy is nearly twice that of regulating power sources. During the dry season, insufficient water inflow leads to a continuous decline in hydropower regulation capacity, and the AGC response level of thermal power units will significantly affect the system's reserve support level, thereby directly affecting the CPS level of inter-provincial interconnection lines. Furthermore, the rapid fluctuations in new energy output make frequency control increasingly difficult.

[0006] The improvement in CPS performance essentially reduces the frequency and power fluctuation range of tie lines, allowing regional power balancing and inter-regional power transmission to be executed as planned. This enhances the overall network's economic efficiency while ensuring system operational safety, which is highly significant for inter-regional interconnected power grids. Therefore, based on the above scenario, while strengthening the construction of load control, pumped storage, and large-scale energy storage stations, it is also essential to deepen and explore ways to improve the performance of the power grid AGC system and optimize strategies. Improving the regulation performance of the AGC system will simultaneously enhance CPS performance and lay a solid foundation for the construction of new power systems.

[0007] For provincial power grids, the response characteristics of each main generating unit participating in AGC (Automatic Guided Control) regulation directly impact the CPS (Continuous Power Response) level of the tie line. The contribution of generating units participating in AGC regulation at different locations to the tie line CPS varies due to factors such as their distance from the tie line, the dispatching strategy of the AGC master station, and their own regulation response capabilities. Meanwhile, current evaluations of generating unit AGC performance primarily consider the power source side, examining indicators such as response rate, response amplitude, and response accuracy. Once the unit meets these response requirements, it is considered sufficient, failing to fully consider the correlation between the unit's AGC regulation response characteristics and the tie line CPS trend. This makes it impossible to propose differentiated methods for improving the AGC response performance of each unit. Summary of the Invention

[0008] The purpose of this invention is to address the shortcomings of the existing technology by providing a method for correlation analysis and performance improvement of unit AGC performance and line CPS indicators. By establishing the correlation, key units are identified and their performance is ranked to improve the AGC performance of the units in a targeted manner, and further improve the CPS level of the tie line.

[0009] To achieve the above objectives, the technical solution of the present invention is: a method for correlation analysis and performance improvement of unit AGC performance and line CPS index, comprising:

[0010] S1. Construction of the correlation between unit automatic generation control response performance and tie line CPS performance: Based on deep learning neural network, a deep learning neural network is constructed with the AGC performance of each unit as input and the tie line CPS performance of the corresponding time period as output. The neural network is trained using historical data of past AGC frequency regulation processes as samples to realize the construction of the functional relationship between unit AGC performance and system CPS performance.

[0011] S2. Contribution and sensitivity of unit AGC to tie line: Based on the relationship between unit output and tie line power deviation, the influence of unit output changes on tie line power deviation adjustment is obtained, and the importance ranking of each unit is derived.

[0012] S3. Improvement of unit AGC adjustment performance: Different strategies are used to analyze the main factors affecting the unit's AGC performance and obtain the overall range of improvement of the unit's AGC performance.

[0013] In one embodiment of the present invention, step S1 is specifically implemented as follows:

[0014] ① Construction of Neural Network Samples: Collect records of AGC (Automatic Generation Control) commands issued by the power grid master station to obtain the AGC performance of the generating units: the regulation capacity, response time, regulation rate, and regulation accuracy of the unit's power output tracking load command process; then, divide the indicators obtained from each command issuance into time periods, with the standard for time period division being the corresponding time of the CPS (Continuous Power Response). If a regulation command spans multiple time periods, the weighting is based on the proportion of the time period occupied by the duration of the command issuance; if the generating unit receives multiple commands within a given time period, the average of the comprehensive indicators at that time is taken as the input for that time period; simultaneously, for generating units that do not participate in AGC regulation during the same time period, their corresponding input is 0; after obtaining the input of the neural network, the corresponding CPS indicator at the given time period is used as the output to obtain a training sample; and so on, to obtain a predetermined number of training samples.

[0015] ② The neural network composed of multilayer perceptrons is trained using training samples; the training samples are divided into training set and test set. After the training converges, the functional relationship between the unit's AGC performance and CPS is established.

[0016] The multilayer perceptron neural network is constructed using a backpropagation (BP) neural network. Its algorithm includes two processes: forward propagation of the signal and backward propagation of the error. Specifically, the error output is calculated from input to output, while the weights and thresholds are adjusted from output to input. During forward propagation, the input signal acts on the output node through the hidden layer, undergoes a nonlinear transformation, and generates the output signal. If the actual output does not match the expected output, the backward propagation process begins. Backpropagation involves propagating the output error back through the hidden layer to the input layer layer by layer, distributing the error to all units in each layer. The error signals obtained from each layer are used as the basis for adjusting the weights of each unit. By adjusting the connection strength between the input node and the hidden layer nodes, the connection strength between the hidden layer nodes and the output node, and the threshold, the error is reduced along the gradient direction. After repeated learning and training, the network parameters corresponding to the minimum error are determined, and training stops. At this point, the trained neural network can automatically process similar input information to produce the information with the minimum output error after nonlinear transformation.

[0017] After repeated training and learning, once the neural network converges, the functional relationship between the unit's AGC performance and the system's CPS performance is established; thus, the correlation between the unit's AGC response capability and the tie-line CPS is obtained.

[0018] In one embodiment of the present invention, step S2 is specifically implemented as follows:

[0019] First, for the regional power system, with the load remaining constant, the output ΔP of the i-th generating unit... i The change will cause a change in the power exchanged between the system and the outside world in this region. Therefore, in the power flow calculation of the power system, the change in unit output and the tie line power deviation ΔP are simplified to the following equation:

[0020]

[0021] By analyzing AGC response data, the changes in tie-line power deviation over different time periods and the corresponding unit output changes were obtained. The changes in tie-line power deviation for each minute were calculated. The corresponding unit output changes at the time points were also calculated. Finally, the changes in tie-line power deviation for each minute and the corresponding unit output changes throughout the day were obtained and the data were integrated. Regression analysis using the least squares method was performed to obtain the corresponding coefficient k for different units. i The impact of adjusting the output of the i-th generating unit on the change in tie-line power deviation was investigated by statistically analyzing the changes in unit output and tie-line power deviation over various time periods, and then fitting the data to obtain the coefficient k. i k i The magnitude of this value reflects the extent to which changes in unit output contribute to changes in tie-line power deviation.

[0022] Secondly, considering the impact of load changes on tie-line power deviation, load data for the corresponding time period is obtained for further analysis, and based on this, sensitivity analysis is performed on changes in unit output and tie-line power deviation. Under steady-state conditions, all loads are treated as a single negative-power unit, and the impact of load fluctuations and changes in unit AGC output on tie-line power deviation is studied, as shown in the following formula:

[0023]

[0024] △P0 represents the overall load change of the entire network at the same time scale, and k0 represents the load change correlation coefficient, which is approximately equal to 1 when network losses are ignored; s i This indicates the sensitivity of the frequency change corresponding to the change in tie line power ΔP in the AGC output response regulation of the i-th unit.

[0025] Next, a sensitivity analysis of the relationship between the unit's AGC response characteristics and frequency changes in the tie-line CPS level is conducted. Specifically, for a typical power network structure, a sensitivity analysis is performed on the relationship between changes in unit output and changes in tie-line power deviation. This analysis examines how much change in unit output will cause in the tie-line power deviation under a given power system operating condition, and describes the local linear relationship between changes in AGC output and changes in tie-line power deviation.

[0026]

[0027] This represents the sensitivity coefficient of frequency change caused by power transmission variation on the tie line, which is affected by the output variation of unit i. Power plants with a larger sensitivity coefficient have a greater control effect on the power of the tie line.

[0028] Finally, by using annual operating data and employing a weighted average method, the sensitivity of each generating unit to changes in tie line power was analyzed, and the contribution of different generating units to changes in tie line CPS power was identified.

[0029] In one embodiment of the present invention, step S3 is specifically implemented as follows:

[0030] For thermal power units, the following strategies are adopted to study and analyze the main factors affecting the AGC performance of thermal power units, and corresponding measures are taken to improve their overall performance:

[0031] Strategy 1: From the perspective of the speed control system and electro-hydraulic servo mechanism, the parameters of PID control include the PID proportional element multiplier K. P PID derivative factor K D PID integral element multiplier K IThe proportional coefficient K2 of the feedforward control system will affect the accuracy of operation control; the function of the electro-hydraulic servo mechanism is to transmit the gate command generated by the regulating system to the gate and change the gate opening. Therefore, by adjusting the control parameters of the speed regulation system and the electro-hydraulic servo mechanism, the quantified range of AGC performance improvement can be obtained.

[0032] Strategy 2: The CCS is the core system for thermal power units to participate in AGC regulation. The controller P in the CCS T loop and N e Feedback controller G of the loop PT (S) and P T Feedforward controller G of the loop ff (S), where S is the Laplace operator, and the expressions for each controller are written as:

[0033]

[0034]

[0035] G ff (S)=k pfd +k dfd / (52S+1);

[0036] Where k p1 and k i1 These are the proportional gain and integral gain of the main steam pressure loop, respectively; k p2 and k i2 These are the proportional gain and integral gain of the main steam pressure loop, respectively; k pdf and k dfd These are the proportional gain and derivative gain of the power loop feedforward controller, respectively; k p1 It will have a significant impact on the adjustment performance of AGC: as k p1 As the rate of increase increases, the adjustment rate index k1 and the response time index k3 will gradually increase. When a predetermined level is reached, the rate of increase of k1 and k3 accelerates rapidly; and as k increases further... p1 As the value increases, the adjustment precision index k2 first increases and then decreases rapidly; k i1 The moderating index of AGC shows a positive correlation, especially with k2; as k... p2 With the increase of k1 and k3, k2 shows a significant increase, while the changes in k1 and k3 become uncertain, exhibiting different trends under different operating conditions, but with relatively small overall amplitudes. Therefore, for k1 and k3, the change in k2 is more significant than the change in k3. p2 The sensitivity of k is not high, but the sensitivity to k2 is relatively high, meaning that k... p2 The improvement of k has a positive impact on the performance of k2; i2The changes in k are relatively small and have a weak correlation with the fluctuations in AGC performance adjustment indicators, which means that k i2 The impact on AGC performance metrics is minimal; k pdf It shows a negative correlation with k2 and a positive correlation with k1; k dfd No significant correlation was found with any of the AGC regulation performance indicators; therefore, k p1 k i1 and k p2 There is a clear correlation between k and the performance indicators of AGC regulation; to improve AGC regulation performance, k is analyzed. p1 k i1 and k p2 The extent to which the improvement affects AGC performance is discussed, and based on its mechanism and operating characteristics, the value of k is given. p1 k i1 and k p2 The adjustable range allows for a quantifiable range of AGC performance improvement.

[0037] Strategy 3: Boiler-side optimization and improvement: The boiler main control strategy is based on load command feedforward. The boiler main control command acts synchronously on the feedwater, fuel, and air volume loops. By setting accurate BM-feedwater, BM-fuel, and BM-air volume f(x) function relationships, the feedforward action ensures that changes in feedwater, coal, and air volume can be achieved in one step during load changes, reaching the predetermined values ​​and ensuring that the unit load also reaches the predetermined values. The feedforward plays a coarse adjustment role, and PID control provides feedback for fine adjustment, ultimately enabling the unit load to reach its set value and stabilizing the main steam temperature and pressure at the target values. Therefore, by analyzing the main influencing factors of the boiler main control on the unit's AGC performance, and then providing its quantitative influencing links and improvement potential, the range of AGC performance improvement due to the improvement of the main control can be obtained.

[0038] Finally, based on strategy one, strategy two, and strategy three, the overall range of improvement that the unit's AGC performance can achieve is obtained.

[0039] Compared with the prior art, the present invention has the following beneficial effects: The present invention establishes the correlation between the automatic generation control response performance of the unit and the CPS performance of the tie line. Based on this, key units are identified and combined with the unit performance ranking, and the AGC performance of the units is improved in a targeted manner, and the CPS level of the tie line is further improved. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of a neural network.

[0041] Figure 2 This paper presents an analysis approach to the impact of unit AGC regulation performance on tie line CPS indicators.

[0042] Figure 3 This is a diagram of the AGC control performance optimization model for thermal power units.

[0043] Figure 4 This is the model structure of the speed control system.

[0044] Figure 5 This is a model of an electro-hydraulic servo mechanism. Detailed Implementation

[0045] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0046] This invention provides a method for correlation analysis and performance improvement of unit AGC performance and line CPS index, including:

[0047] S1. Construction of the correlation between unit automatic generation control response performance and tie line CPS performance: Based on deep learning neural network, a deep learning neural network is constructed with the AGC performance of each unit as input and the tie line CPS performance of the corresponding time period as output. The neural network is trained using historical data of past AGC frequency regulation processes as samples to realize the construction of the functional relationship between unit AGC performance and system CPS performance.

[0048] S2. Contribution and sensitivity of unit AGC to tie line: Based on the relationship between unit output and tie line power deviation, the influence of unit output changes on tie line power deviation adjustment is obtained, and the importance ranking of each unit is derived.

[0049] S3. Improvement of unit AGC adjustment performance: Different strategies are used to analyze the main factors affecting the unit's AGC performance and obtain the overall range of improvement of the unit's AGC performance.

[0050] The following is a detailed implementation process of the present invention.

[0051] This invention discloses a method for correlation analysis and performance improvement of unit AGC performance and line CPS index, comprising the following steps:

[0052] (I) A method for constructing the correlation between the automatic generation control response performance of a generating unit and the performance of the tie-line CPS.

[0053] Based on deep learning neural networks, a deep learning neural network is constructed with the AGC performance of each generating unit as input and the CPS performance of the tie line during the corresponding time period as output. Historical data from past AGC frequency regulation processes are used as samples to train the neural network, thereby establishing a functional relationship between the generating unit AGC performance and the system CPS performance. The specific steps are as follows:

[0054] Construction of Neural Network Samples. Records of AGC (Automatic Generation Control) commands issued by the power grid master station are collected to obtain the AGC performance of the generating units: regulation capacity, response time, regulation rate, and regulation accuracy during the process of tracking load commands with unit power output. Then, the indicators obtained from each command are divided into time periods, with the standard for time division being the corresponding time of the CPS (Controlled Power Statistical Point). If a regulation command spans multiple time periods, the weighting is based on the proportion of the command's duration within each time period. If a unit receives commands multiple times within a given time period, the average of the comprehensive indicators for that time period is taken as the input for that time period. Simultaneously, for units not participating in AGC regulation during the same time period, their corresponding input is 0. After obtaining the input to the neural network, the corresponding CPS indicator at each time period is used as the output, resulting in a training sample. A sufficiently large number of training samples are obtained using a similar method.

[0055] The neural network, composed of a multilayer perceptron, was trained using training samples. The training samples were divided into a training set and a test set. After training converged, a functional relationship between the unit's AGC performance and CPS was established.

[0056] The multilayer perceptron neural network is constructed using a backpropagation (BP) neural network. Its algorithm includes two processes: forward propagation of the signal and backward propagation of the error. That is, the error output is calculated from input to output, while the weights and thresholds are adjusted from output to input. During forward propagation, the input signal acts on the output node through the hidden layer, undergoes a nonlinear transformation, and generates the output signal. If the actual output does not match the expected output, the backward propagation process begins. Backpropagation propagates the output error layer by layer back to the input layer, distributing the error among all units in each layer. The error signals obtained from each layer are used as the basis for adjusting the weights of each unit. By adjusting the connection strength between the input node and the hidden layer nodes, the connection strength between the hidden layer nodes and the output node, and the threshold, the error is reduced along the gradient direction. After repeated learning and training, the network parameters (weights and thresholds) corresponding to the minimum error are determined, and training stops. At this point, the trained neural network can automatically process similar input information to minimize the output error after nonlinear transformation. A schematic diagram of the neural network structure is shown below. Figure 1 As shown:

[0057] W in the diagram jk l b represents the weights on the connections from the k neurons in layer l-1 to the j-th neuron in layer l; j l This represents the deviation of j neurons in layer l; a j l This represents the activation value of j neurons in layer l. The activation formula for a single neuron can be expressed as:

[0058]

[0059] Where sum(l-1) represents the number of neurons in layer (l-1), and σ is the activation function relationship.

[0060] The weights w, bias b, and activation value a are represented by matrices. The values ​​of the l-th layer are w, b, and a, respectively. l b l a l Equation (1) can be rewritten using the single neuron activation formula of the matrix as follows:

[0061] a l =σ(w l a l-1 +b l (2)

[0062] For equation (2), the weighted input z of a single neuron can be proposed. j l :

[0063]

[0064]

[0065] The loss function can be set in various forms, including:

[0066]

[0067] Equation (5) can be written in matrix form:

[0068]

[0069] Define the error on the j-th neuron of layer l. for:

[0070]

[0071] By deriving equation (7), we can obtain:

[0072]

[0073] Single neuron error Relationship with the next layer (l+1 layer):

[0074]

[0075] Relationship between error and weight:

[0076]

[0077] Similarly, the relationship between error and deviation b can be obtained:

[0078]

[0079] From the above formula, we can conclude that after one forward propagation, the partial derivatives of C with respect to each w and b can be quickly solved by the error of the output layer. By adding the bias to each w and b, the error can be reduced, thus making the convergence direction of the neural network consistent with the desired direction.

[0080] After repeated training and learning, once the neural network converges, the functional relationship between the unit's AGC performance and the system's CPS performance is established. Thus, the correlation between the unit's AGC response capability and the tie-line CPS is obtained.

[0081] (II) Contribution and sensitivity of unit AGC to tie-line participation

[0082] Furthermore, a method for improving the CPS performance of tie lines based on the improvement of unit AGC performance specifically includes:

[0083] This study investigates the relationship between unit output and tie-line power deviation, further examining the impact of unit output changes on tie-line power deviation adjustments, and deriving a ranking of the units' importance. Based on this, and according to the importance ranking of different units, the study considers the performance improvement potential of each unit and investigates AGC unit performance improvement strategies and corresponding implementation methods. Since power balance involves multiple aspects such as unit, load, and renewable energy output fluctuations, the concept of influencing factors is introduced. The research approach for the relationship between unit AGC response ranking and tie-line CPS changes is as follows: Figure 2 As shown:

[0084] For a regional power system, with the load remaining constant, the output ΔP of the i-th generating unit... i The change will cause a change in the power exchanged between the system and the outside world in this region. Therefore, in the power flow calculation of the power system, the change in unit output and the power deviation ΔP of the tie line can be simplified to the following equation:

[0085]

[0086] By analyzing AGC response data, the changes in tie-line power deviation over different time periods and the corresponding changes in unit output at those times were obtained. The research approach was as follows: The change in tie-line power deviation was calculated for each minute; the changes in unit output at the corresponding times were calculated; the changes in tie-line power deviation per minute and the changes in unit output during the same period throughout the day were obtained, and the data were integrated; regression analysis was performed using the least squares method to obtain the corresponding coefficient k for different units. iThe impact of adjusting the output of the i-th generating unit on the change in tie-line power deviation was investigated. This was achieved by statistically analyzing the changes in generating unit output and tie-line power deviation over various time periods, and then fitting the data to derive the coefficient k. i k i The magnitude of the value reflects the extent to which changes in unit output contribute to changes in tie-line power deviation.

[0087] Furthermore, considering the impact of load changes on tie-line power deviation, load data for the corresponding time period is obtained for further analysis, and based on this, sensitivity analysis is performed on changes in unit output and tie-line power deviation. Under steady-state conditions, all loads can be considered as a single negative-power unit, and the impact of load fluctuations and changes in unit AGC output on tie-line power deviation is studied, as shown in the following equation:

[0088]

[0089] △P0 represents the overall load change of the entire network at the same time scale, and k0 represents the load change correlation coefficient, which can be approximated as 1 if network losses are ignored; s i This represents the sensitivity of the frequency change corresponding to the change in tie line power ΔP when the AGC output response regulation of the i-th unit is adjusted.

[0090] Furthermore, a sensitivity analysis was conducted on the relationship between the unit's AGC response characteristics and frequency changes in the tie-line CPS level. Specifically, for a typical power network structure, a sensitivity analysis was performed on changes in unit output and changes in tie-line power deviation. This analysis examines how much change in unit output will cause in the tie-line power deviation under a given power system operating condition, and describes the local linear relationship between changes in AGC output and changes in tie-line power deviation.

[0091]

[0092] This represents the sensitivity coefficient of frequency change caused by power transmission variation on the tie line, which is affected by the output variation of unit i. Power plants with a larger sensitivity coefficient have a greater control effect on the power of the tie line.

[0093] Furthermore, by using annual operating data and employing a weighted average method, the sensitivity of each unit to changes in tie line power is analyzed, and the contribution of different units to changes in tie line CPS power is identified. For specific units, the following methods can be adopted for optimization to further improve the unit's AGC performance.

[0094] (III) Improved AGC regulation performance of the unit

[0095] The following are methods to improve the AGC (Automatic Guided Vehicle) regulation performance of power units. Taking thermal power units as an example, many factors affect their AGC regulation performance, and the influencing factors vary depending on the type of unit. The main factors affecting AGC performance include: insufficient or slow output of major equipment, poor quality of important AGC-related signals, poor quality of lower-level automatic regulation, insufficient or inappropriate boiler feedforward after AGC command changes, and delayed turbine response after AGC command changes. The main influencing factors on the AGC control performance of thermal power units are as follows: Figure 3 As shown:

[0096] For thermal power units, changing the proportional and integral parameters in the turbine main control regulator can effectively increase the regulator's output speed, thereby increasing the execution rate after the unit load command is issued, and thus enabling the unit load command to quickly track changes in AGC commands. For example... Figure 4 , 5 For a certain model of speed control system shown, it also has a significant impact on AGC performance.

[0097] Figure 5 In the diagram, parameter Δω represents the difference between the actual rotational speed and the base rotational speed, and P... REF P is the load setpoint. M1 For the mechanical power output of the steam turbine, P E For the generator's electrical power output, P CV For the valve opening command, T1 is the time constant of the speed measurement circuit, K is the speed deviation amplification factor, K2 is the proportional coefficient, S is the Laplace operator, and K... P K is the proportional component factor in a PID controller. D K is the multiplier of the PID derivative element. I This is the multiplier for the integral element of the PID controller.

[0098] P CV For the valve opening command, P GV For the valve opening, T C T is the shut-off time constant of the hydraulic actuator. O K is the hydrator start-up time constant. EHC K is the overspeed shutdown time coefficient. EHO EHC is the overspeed start-up time constant. close EHC is used to close the limiter at excessive speed. open To enable amplitude limiting at excessive speed, P MAX P represents the maximum stroke of the hydraulic actuator or the maximum opening of the control valve. MIN T2 is the minimum stroke of the hydraulic actuator or the minimum opening of the control valve, and K is the stroke feedback time of the hydraulic actuator. P K is the proportional component factor in a PID controller. D K is the multiplier of the PID derivative element. I The integral factor of the PID controller is the PID multiplier.max The maximum amplitude limit output by the PID module, PID min This is the minimum value of the amplitude limit output by the PID module.

[0099] according to Figure 3 AGC influencing factors and Figure 4 , Figure 5 Regarding the speed control system, this project plans to adopt the following strategies to study and analyze the main factors affecting the AGC performance of thermal power units, and take effective measures to improve its overall performance.

[0100] Strategy 1: From the perspective of the speed control system and electro-hydraulic servo mechanism, the parameters of PID control (including K) P K D K I Both the proportional gain and the feedforward coefficient K2 affect the accuracy of the operation control. For PID control, it is known that proportional regulation has a fast adjustment effect, but it will produce steady-state error, meaning that the proportional element cannot eliminate the error. The integral element eliminates steady-state error by integrating the error. The derivative element is mainly used to suppress the response error; differentiating the system error will avoid overshoot. The feedforward control system is a control system that works according to the principle of compensation based on the change of disturbance or setpoint. Its characteristic is that when a disturbance occurs, before the controlled variable changes, it controls according to the magnitude of the disturbance to compensate for the influence of the disturbance on the controlled variable. When used properly, the feedforward control system can eliminate the disturbance of the controlled variable in its infancy, so that the controlled variable will not deviate due to the disturbance or change of setpoint. Therefore, the feedforward coefficient will affect the system deviation. From the perspective of adjustment speed and adjustment accuracy, the setting of the control parameters of the speed regulation system will inevitably affect the adjustment error, and thus affect the adjustment accuracy and adjustment speed. The function of the electro-hydraulic servo mechanism is to transmit the adjustment gate command generated in the adjustment system to the adjustment gate, changing the opening of the adjustment gate.

[0101] Strategy Two: The CCS (Control System) is the core system for thermal power units participating in AGC (Automatic Guided Control) regulation, characterized by strong nonlinearity and strong coupling. The controller P in the CCS... T loop and N e Feedback controller G of the loop PT (S) and P T Feedforward controller G of the loop ff (S), where S is the Laplace operator, and the expressions for each controller can be written as:

[0102]

[0103]

[0104] G ff (S)=kpfd +k dfd / (52S+1);

[0105] Where k p1 and k i1 These are the proportional gain and integral gain of the main steam pressure loop, respectively; k p2 and k i2 These are the proportional gain and integral gain of the main steam pressure loop, respectively; k pdf and k dfd These are the proportional gain and derivative gain of the power loop feedforward controller, respectively. p1 It will have a significant impact on the adjustment performance of AGC: as k p1 As k increases, k1 and k3 (k1 refers to the adjustment rate index, and k3 refers to the response time index, the same below) will gradually increase. When they reach a certain level, the rate of increase of k1 and k3 will accelerate rapidly. And as k... p1 As the value increases, k2 (the adjustment precision index, the same below) first increases and then decreases rapidly. i1 The moderating index of AGC shows a positive correlation, especially with k2. As k... p2 With the increase of k1 and k3, k2 shows a significant increase, while the changes in k1 and k3 become uncertain, exhibiting different trends under different operating conditions, but the overall magnitude of change is relatively small. Therefore, for k1 and k3, the relationship between k1 and k3 and k2 is relatively stable. p2 The sensitivity of k is not high, but the sensitivity to k2 is relatively high, meaning that k... p2 The improvement of k has a positive impact on the performance of k2. i2 The changes in k are relatively small and have a weak correlation with the fluctuations in AGC performance adjustment indicators, which means that k i2 The impact on AGC performance metrics is minimal. pdf It shows a certain degree of negative correlation with k2 and a certain degree of positive correlation with k1. dfd No significant correlation was found with any of the AGC regulation performance indicators. Therefore, it is not difficult to see that k p1 k i1 and k p2 There is a clear correlation between k and the AGC regulation performance indicators. To improve AGC regulation performance, this project, based on the above analysis, further conducts in-depth quantitative analysis of k. p1 k i1 and k p2 The extent to which the improvement affects AGC performance is discussed, and based on its mechanism and operating characteristics, the value of k is given. p1 k i1 and k p2 The range is further adjustable, thus obtaining a quantified range of AGC performance improvement.

[0106] Strategy 3: Boiler-Side Optimization and Improvement. The boiler main control strategy is primarily based on load command feedforward. The boiler main control commands synchronously act on the feedwater, fuel, and airflow loops. By setting accurate BM-feedwater, BM-fuel, and BM-airflow f(x) functional relationships, during load changes, the feedforward action ensures that changes in feedwater, coal, and airflow can be achieved almost instantly to predetermined values, thus guaranteeing that the unit load also adjusts accordingly. Feedforward acts as a coarse adjustment, while PID control provides feedback for fine adjustment, ultimately ensuring the unit load reaches its setpoint and the main steam temperature and pressure stabilize at target values. Therefore, this study aims to analyze the main influencing factors of the boiler main control on the unit's AGC performance, and then quantify the influencing links and improvement potential, thereby determining the range of AGC performance improvement resulting from main control improvements.

[0107] Finally, based on the above three strategies, the overall range of improvement that the unit's AGC performance can achieve is obtained.

[0108] Specific application examples of the method of this invention.

[0109] 1.1 Analysis of the Relationship between AGC and CPS Indicators

[0110] During the AGC control process, different AGC units participate at different times, and the regulation performance of each unit also varies. If the overall performance of the units participating in AGC regulation is better within a certain period, it means that the system can eliminate frequency deviations and regional control deviations more quickly. In this case, the units participating during that period are more beneficial for CPS assessment. Therefore, studying the relationship between unit performance and CPS can further analyze the impact of unit performance improvement on CPS improvement. Based on the above analysis, this study uses historical operating data of AGC units in a local power grid and power grid CPS indicators to study the relationship between the two, and further analyzes the degree of improvement of system CPS performance by AGC performance.

[0111] Collect records issued by the power grid in a certain area to obtain the AGC performance of the generating units: the response time, regulation rate and regulation accuracy of the unit's power output tracking load command process, and calculate the corresponding comprehensive performance indicators.

[0112] Next, the indicators obtained from each issued command are divided into time periods, with the standard for time division being the corresponding time of CPS. The data is divided into different time periods, and the generating units participating in the adjustment within each time period are relatively fixed. Then, the relationship between generating unit performance and CPS standards is studied within each time period.

[0113] 1.2 Construction of the functional relationship between unit performance and CPS index

[0114] The following table shows some data obtained in June 2023:

[0115] Table 1

[0116]

[0117] In the table above, a unit performance of 0 indicates that the unit did not participate in AGC regulation during that period. Using a time period (e.g., 10 minutes) as the cycle, calculate the average performance K of the i-th unit when it participates in regulation during that period. i and CPS1 average Will With unit performance as the dependent variable and unit performance as the independent variable, a linear regression method is used to fit the relationship between unit performance and... The functional relationship is used to obtain the unit performance of all N units involved in the regulation during that time period. Functional relationship:

[0118]

[0119] Taking 15 generating units that participated in the regulation during a certain period in June 2023 as an example, the relationship coefficients between CPS1 and unit performance are shown in Table 2 below:

[0120] Table 2

[0121]

[0122] A larger coefficient indicates a stronger correlation between the unit and the CPS index; improving the unit's performance is beneficial to improving the CPS index. The performance index K of the i-th unit... i The degree of improvement in the CPS index due to the increase in the coefficient can also be represented by this coefficient.

[0123] 2. Research on the ranking method of the comprehensive performance of AGC units and the measurement method of the degree of improvement of AGC unit performance on the CPS performance of the system.

[0124] 2.1 Ranking Method for Overall Performance of AGC Units

[0125] A comprehensive analysis is conducted on the regulation performance indicators of the AGC unit (such as regulation rate K1, response time K2, and regulation accuracy K3) to obtain the comprehensive performance index of the AGC unit. The comprehensive regulation performance index is expressed by a weighted sum or product of k1, k2, and k3. This paper adopts the product method, i.e.:

[0126] K = k1 × k2 × k3

[0127] Since the comprehensive index is a fluctuating quantity, statistical techniques are further used to obtain its mean and variance. Then, the mean and variance are weighted to obtain the comprehensive performance value of the unit's AGC performance. Based on the magnitude of this comprehensive performance value, the AGC performance of the units can be ranked, with those ranking higher indicating better comprehensive performance. Table 3 below shows some unit performance indices and variances as of June 1, 2023:

[0128] Table 3

[0129] Banling Pumped Storage Power Plant Unit #1 2.939002 0.919971 Shuikou Hydropower Plant Unit #4 3.263718 0.461681 Shuikou Hydropower Plant Unit #1 0.606826 0.147608 Chitan Hydropower Plant Unit #1 0.663057 1.029175 Shizhen Thermal Power Plant Unit #2 2.533 0.767128 Ningde Thermal Power Plant Unit #2 1.237809 1.047728 Fujian Xin'ao Thermal Power Plant, Unit #2 4.167602 7.291083 Chitan Hydropower Plant Unit #3 1.852119 2.851092 Ningde Thermal Power Plant Unit #4 0.843348 0.462801 Houshi Thermal Power Plant Unit #2 2.071111 1.38321 Houshi Thermal Power Plant Unit #3 1.3242 1.084484 Nanpu Thermal Power Plant Unit #4 2.666471 1.59957 Kemen Thermal Power Plant Unit #3 1.61832 0.946442

[0130] Variance describes the degree of fluctuation in unit performance. A large variance indicates that the unit's performance is unstable. Therefore, the ratio of the mean to the variance can be used as a comprehensive indicator of unit performance. The unit performance ranking is shown in Table 4 below:

[0131] Table 4

[0132] Shuikou Hydropower Plant Unit #4 7.06921 Shuikou Hydropower Plant Unit #1 4.111053 Shizhen Thermal Power Plant Unit #2 3.301928 Banling Pumped Storage Power Plant Unit #1 3.194667 Ningde Thermal Power Plant Unit #4 1.822271 Kemen Thermal Power Plant Unit #3 1.709899 Nanpu Thermal Power Plant Unit #4 1.666992 Houshi Thermal Power Plant Unit #2 1.497322 Houshi Thermal Power Plant Unit #3 1.221041 Ningde Thermal Power Plant Unit #2 1.181422 Chitan Hydropower Plant Unit #3 0.649617 Chitan Hydropower Plant Unit #1 0.644261 Fujian Xin'ao Thermal Power Plant, Unit #2 0.571603

[0133] 2.2 Methods for measuring the degree to which improved AGC unit performance contributes to the improvement of system CPS performance

[0134] The first part has established the functional relationship between different unit performance indicators and CPS:

[0135]

[0136] By assessing the sensitivity of the CPS (Common Performance Status) index to the overall performance indicators of AGC (Automatic Gauge Control) units, the strength of the correlation between the overall performance of AGC units and the CPS index is obtained. For AGC units with high sensitivity, changes in their performance have a significant impact on the CPS index. Improving the regulation performance of these units can yield better potential for CPS performance improvement. Finally, by analyzing the relationship between the overall performance of the units and the sensitivity of the CPS, and the performance improvement potential of each unit, the degree to which improvements in AGC unit performance contribute to the improvement of system CPS performance is measured.

[0137] If a linear regression method is used to establish the functional relationship between the two, the larger the coefficient, the greater the influence of the generator unit on the CPS index. Further research can be conducted on the potential for improvement in generator unit performance. By substituting the generator unit performance into the established function, the corresponding potential for improvement in CPS can be obtained. Some data from June 2023 are shown in Table 5 below:

[0138] Table 5

[0139]

[0140] The coefficient of X Variable 1 is 1.279, indicating that for every unit improvement in unit performance, the corresponding increase in CPS1 is approximately 1.279. To ensure the rationality of this regression method, it is necessary to find relevant data characteristics on a longer time scale and establish a functional relationship to measure the improvement in CPS caused by the improvement in unit performance.

[0141] 3. Research on methods to improve the regulation performance of AGC units

[0142] 3.1 Technical Route Analysis

[0143] Many factors influence CPS performance, including generator AGC performance, generation plans, grid configuration, abnormal power reception curves, frequency during special periods, and daily load variations. Since these factors are largely beyond human control, and generator AGC performance is a major factor affecting CPS performance, improving generator regulation performance is crucial for CPS improvement. This project proposes the following strategies to analyze the main factors affecting thermal power unit AGC performance and implement effective measures to improve its overall performance.

[0144] Methods for improving AGC (Automatic Generative Control) regulation performance can be broadly categorized into two types: those based on unit equipment modification and upgrades, and those based on optimizing unit operation control strategies. The basic idea behind AGC regulation performance improvement methods based on unit equipment modification and upgrades is to enhance the unit's AGC regulation performance by modifying and upgrading the unit itself or its auxiliary equipment. The basic idea behind methods based on optimizing unit operation control strategies is to improve AGC regulation performance without altering the current hardware of the unit and control system, but by adopting superior operation control strategies.

[0145] This proposal suggests a novel method for improving the AGC (Automatic Generative Control) performance of generator units based on sensitivity analysis and parameter optimization. First, a perturbation-based method for calculating generator AGC performance is presented. Then, sensitivity analysis is used to identify key factors and parameters that significantly impact AGC performance. Finally, a genetic algorithm is employed to optimize these key parameters, thereby enhancing the generator AGC regulation performance.

[0146] 3.2 Identification of key elements based on sensitivity analysis

[0147] S xi The unit's overall performance index K represents the parameter x. i The sensitivity. If the overall performance y of the unit is related to the parameters [x1, x2, ..., x...] n The relationship can be represented as y = f(x1, x2, ..., x). nAt the current operating point, the performance index y is relative to the parameter x. i The expression for calculating sensitivity is:

[0148]

[0149] In the above formula, if the analytical function expressions of the system performance y and parameter x are known, the corresponding relationship can be obtained directly by differentiating the functional expressions of y and x. Because thermal power units have many control links and complex structures, and are a hybrid system of linear and nonlinear links, the functional relationships between all parameters and outputs (performance indicators) are very complex, making direct derivation of these relationships difficult. Therefore, by building a model of the control system, sensitivity is calculated using simulation, and key links and parameters are obtained based on their magnitude. The specific process is as follows:

[0150] (1) Construct a simulation model of the thermal power unit control system and set the current values ​​of all parameters x c,i (i = 1, 2, ..., n) are set to their normal values, and each parameter x is... i The allowable range of variation is [x i,min ,x i,max ];

[0151] (2) Load command increased by ΔN E Calculate the comprehensive performance index based on the response curve and denote it as K. base i = 1;

[0152] (3) Parameter x i Increase the current value by 5% × (x) i,max -x i,min The load command increases by ΔN. E The comprehensive performance index is obtained from the response curve and denoted as K. i ;

[0153] (4) Calculate sensitivity Right now:

[0154]

[0155] (5) i = 1 + 1; if i equals n + 1, the calculation ends; otherwise, proceed to (4);

[0156] Based on the above, the sensitivity of each parameter under different values ​​is further calculated, and the average sensitivity is obtained; based on the sensitivity threshold, the key links and key parameters are identified.

[0157] A unit model and control loop were constructed. The boiler and turbine main controllers employed PI control, the boiler feedforward control was a proportional-derivative feedforward control based on power deviation, and the turbine speed regulation system's DEH (decentralized feedback) control used proportional feedback control. A 300MW unit #1 from a power plant was selected, and parameter values ​​for the turbine, boiler, and DEH models were obtained through parameter identification. The sensitivities of the relevant parameters were calculated using sensitivity analysis, as shown in Table 6.

[0158] Table 6 Sensitivity coefficients of controller parameters to overall regulation performance

[0159]

[0160] The sensitivity coefficients, from largest to smallest, are K i1 >K pf >K p1 >K p2 >K df >K i2 >K p3 .in, and The sensitivity coefficient is relatively small, indicating that adjusting these parameters has little effect on improving system performance. and These parameters have a significant impact on the regulation performance of thermal power units. By setting these parameters appropriately, the unit's performance can be effectively improved. Therefore, optimization is possible. and Improve the AGC load response performance of thermal power units.

[0161] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for correlation analysis and performance improvement of unit AGC performance and tie-line CPS index, characterized in that, include: S1. Construction of the correlation between the unit's automatic generation control response performance and the tie-line CPS performance: Based on a deep learning neural network, a deep learning neural network is constructed with the AGC performance of each unit as input and the tie-line CPS performance of the corresponding time period as output. The neural network is trained using historical data of the AGC frequency regulation process as samples to realize the construction of the functional relationship between the unit's AGC performance and the tie-line CPS performance. S2. Unit AGC Contribution and Sensitivity Analysis to Tie Line: Based on the relationship between unit output and tie line power deviation, the influence of unit output changes on tie line power deviation adjustment is obtained, and the importance ranking of each unit is derived; the specific implementation is as follows: First, for the regional power system, with the load remaining constant, the output ΔP of the i-th generating unit... i The change will cause a change in the power exchange capacity of the regional power system. Therefore, in the power flow calculation of the power system, the change in unit output and the tie-line power deviation ΔP are simplified to the following equation: By analyzing AGC response data, the changes in tie-line power deviation over different time periods and the corresponding unit output changes were obtained. The changes in tie-line power deviation for each minute were calculated. The corresponding unit output changes at the time points were also calculated. Finally, the changes in tie-line power deviation for each minute and the corresponding unit output changes throughout the day were obtained and the data were integrated. Regression analysis using the least squares method was performed to obtain the corresponding coefficient k for different units. i The impact of adjusting the output of the i-th generating unit on the change in tie-line power deviation was investigated by statistically analyzing the changes in unit output and tie-line power deviation over various time periods, and then fitting the data to obtain the coefficient k. i k i The magnitude of this value reflects the extent to which changes in unit output contribute to changes in tie-line power deviation. Secondly, considering the impact of load changes on tie-line power deviation, load data for the corresponding time period is obtained for further analysis, and based on this, sensitivity analysis is performed on changes in unit output and tie-line power deviation. Under steady-state conditions, all loads are treated as a single negative-power unit, and the impact of load fluctuations and changes in unit AGC output on tie-line power deviation is studied, as shown in the following formula: △P0 represents the overall load change of the entire network at the same time scale, and k0 represents the load change correlation coefficient, which is approximately equal to 1 when network losses are ignored; s i This indicates the sensitivity of the frequency change corresponding to the change in the power deviation ΔP of the tie line in the AGC output response regulation of the i-th unit; Next, a sensitivity analysis of the relationship between the unit's AGC response characteristics and frequency changes in the tie-line CPS level is conducted. Specifically, for a typical power network structure, a sensitivity analysis is performed on the relationship between changes in unit output and changes in tie-line power deviation. This analysis examines how much change in unit output will cause in the tie-line power deviation under a given power system operating condition, and describes the local linear relationship between changes in AGC output and changes in tie-line power deviation. This represents the sensitivity coefficient of frequency change caused by power transmission variation on the tie line, which is affected by the power output variation of unit i. Power plants with a larger sensitivity coefficient have a greater effect on the power regulation of the tie line. Finally, by using annual operating data and employing a weighted average method, the sensitivity of each unit to changes in tie line power deviation was obtained, and the contribution of different units to changes in tie line CPS power was identified. S3. Unit AGC Performance Improvement: Different strategies are used to analyze the main factors affecting the unit's AGC performance to obtain the overall range that the unit's AGC performance can be improved. The overall range that the unit's AGC performance can be improved is then substituted into the functional relationship between the unit's AGC performance and the tie line CPS performance to obtain the value that the tie line CPS performance can be improved.

2. The method for correlation analysis and performance improvement of unit AGC performance and tie-line CPS index according to claim 1, characterized in that, Step S1 is implemented as follows: Construction of Neural Network Samples: Records of AGC (Automatic Generation Control) commands issued by the power grid master station are collected to obtain the AGC performance of the generating units: the regulation capacity, response time, regulation rate, and regulation accuracy of the unit's power output tracking load commands. Then, the indicators obtained from each command are divided into time segments, with the time segmentation standard being the time corresponding to the CPS (Continuous Power Response). If a regulation command spans multiple time segments, the weighting is based on the proportion of the command's duration within each segment. If a unit receives multiple commands within a given time segment, the average of the comprehensive indicators at that time is used as the input for that time segment. Simultaneously, for units not participating in AGC regulation during the same time period, their corresponding input is 0. After obtaining the input to the neural network, the corresponding CPS indicator at that time is used as the output to obtain a training sample. This process is repeated to obtain a predetermined number of training samples. The neural network composed of multilayer perceptrons is trained using training samples; the training samples are divided into training set and test set, and after the training converges, the functional relationship between the unit's AGC performance and CPS is established.

3. The method for correlation analysis and performance improvement of unit AGC performance and tie-line CPS index according to claim 1, characterized in that, Step S3 is implemented as follows: For thermal power units, the following strategies are adopted to study and analyze the main factors affecting the AGC performance of thermal power units, and corresponding measures are taken to improve their overall performance: Strategy 1: From the perspective of the speed control system and electro-hydraulic servo mechanism, the parameters of PID control include the PID proportional element multiplier K. P PID derivative factor K D PID integral element multiplier K I And the proportional coefficient K2 of the feedforward control system; the function of the electro-hydraulic servo mechanism is to transmit the gate command generated by the regulating system to the gate and change the gate opening. Therefore, by adjusting the control parameters of the speed regulating system and the electro-hydraulic servo mechanism, the quantitative range of AGC performance improvement can be obtained. Strategy Two: CCS is the core system for thermal power units to participate in AGC regulation. The CCS contains... Feedback controller of the loop , Feedback controller of the loop , Feedforward controller of the loop The expression is written as: ； ； ； Where S is the Laplace operator, and They are respectively The proportional gain and integral gain of the loop; and They are respectively The proportional gain and integral gain of the loop; and These are the proportional gain and derivative gain of the power loop feedforward controller, respectively. It will have a significant impact on the adjustment performance of AGC: with As the rate of increase increases, the adjustment rate index k1 and the response time index k3 will gradually increase. When a predetermined level is reached, the rate of increase of k1 and k3 will accelerate rapidly. As the value increases, the adjustment precision index k2 first increases and then decreases rapidly. It is positively correlated with the moderating index of AGC, especially with k2; as As the coefficients increase, k2 shows a significant rise, while the changes in k1 and k3 become uncertain, exhibiting different trends under different operating conditions, but with relatively small overall amplitudes. Therefore... The sensitivity to k1 and k3 is not high, but the sensitivity to k2 is relatively high, meaning that... The improvement has a positive impact on the performance of k2; The changes in these parameters have little correlation with fluctuations in AGC performance indicators, indicating that... The impact on AGC performance metrics is minimal. It is negatively correlated with k2 and positively correlated with k1; No significant correlation was found with any of the AGC performance metrics; therefore, , and There is a clear correlation between AGC performance indicators and other performance metrics; to improve AGC performance, analysis is conducted... , and The extent to which improvements affect AGC performance is discussed, and based on its mechanism and operating characteristics, the following is given: , and The adjustable range allows for a quantifiable range of AGC performance improvement. Strategy 3: Boiler-side optimization and improvement: The boiler main control strategy is based on load command feedforward. The boiler main control command acts synchronously on the feedwater, fuel, and air volume loops. By setting accurate BM-feedwater, BM-fuel, and BM-air volume f(x) functional relationships, the feedforward action ensures that the feedwater, coal, and air volume change to predetermined values ​​during load changes, guaranteeing that the unit load also reaches the predetermined values. The feedforward acts as a coarse adjustment, and PID control acts as feedback for fine adjustment, ultimately enabling the unit load to reach its set value and stabilizing the main steam temperature and pressure at target values. Therefore, by analyzing the main influencing factors of boiler main control on the unit's AGC performance, we can quantify the influencing links and improvement potential, thus obtaining the range of AGC performance improvement brought about by boiler main control improvements. Finally, based on strategy one, strategy two, and strategy three, the overall range of improvement that the unit's AGC performance can achieve is obtained.

Images