Power prediction control method, device, equipment and medium
By adopting a power prediction control method based on a controlled autoregressive sliding average model in the thermal power generator set, the problem that traditional control systems are difficult to deal with large delays and inertia links is solved, and simpler and more effective power control is achieved.
Patent Information
- Application Number
- CN202510149987.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-10
- Publication Date
- 2025-05-30
AI Technical Summary
In complex thermal power generator sets, traditional PID single-loop control systems are difficult to effectively control the links of large delays and inertia, resulting in complex control and inconvenient use.
The power prediction control method based on the controlled autoregressive sliding average model is adopted to obtain the local optimal solution through rolling finite time domain optimization, and the prediction model is dynamically corrected to adapt to the actual error.
It provides a convenient power prediction and control method, simplifies the control process at the industrial site, and improves the robustness of the control effect and practical application value.
Smart Images
Figure CN120065724A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of power control, and particularly relates to a power prediction control method, device, equipment and medium. Background Art
[0002] The situation in industrial sites is generally complex. There are many links with large time delays and large inertias in thermal power generating units. It is difficult for general PID single-loop control systems to control such links well. In the past, control structures such as feedforward and cascade were generally used to control the system using multiple PIDs. This control method is relatively complex. Among them, when changing the control loop and using multiple PIDs for control, multiple parameters need to be tuned, which brings great inconvenience to the use in industrial sites.
[0003] In summary, there is an urgent need for a convenient power prediction control technical solution. Summary of the Invention
[0004] In view of the above problems, the present disclosure provides a power prediction control method, device, equipment and medium.
[0005] In a first aspect, a power prediction control method, the method includes:
[0006] Establish a prediction model based on a controlled autoregressive moving average model;
[0007] Based on the prediction model, adopt rolling finite-time domain optimization to obtain a local optimal solution and use it for power control;
[0008] According to the real-time result of power control, dynamically correct the error between the prediction model and the actual situation.
[0009] Further, adopting rolling finite-time domain optimization to obtain a local optimal solution and use it for power control includes:
[0010] At the current moment, based on the control effect in the finite-time domain and calculate the control scheme in the finite-time domain;
[0011] Obtain the local optimal solution from the control scheme and send the control quantity at the current moment to the power system;
[0012] At the next sampling moment, the finite-time domain is shifted backward and the local optimal solution is recalculated.
[0013] Further, the prediction model adopts a second-order system formed by connecting two inertia links of the preset gain and inertia time of the controlled object in series.
[0014] Further, dynamically correcting the error between the prediction model and the actual situation includes:
[0015] Judge the error between the prediction model and the actual situation by overshoot and adjustment time.
[0016] Further, dynamically correct the error between the prediction model and the actual situation, including:
[0017] Dynamically correct the error between the prediction model and the actual situation by prediction horizon, control horizon and softening coefficient.
[0018] Further, when the proportional coefficient of the prediction model is out of balance, increase the prediction horizon, decrease the control horizon, and increase the softening coefficient.
[0019] Further, when the inertia time of the prediction model is out of balance, decrease the control horizon and increase the softening coefficient.
[0020] In a second aspect, a power prediction control device includes: a model establishment unit, a control unit, and a correction unit;
[0021] The model establishment unit is used to establish a prediction model based on the controlled auto-regressive moving average model;
[0022] The control unit is used to obtain a local optimal solution based on the prediction model by using rolling finite horizon optimization and is used for power control;
[0023] The correction unit is used to dynamically correct the error between the prediction model and the actual situation according to the real-time result of power control.
[0024] In a third aspect, an electronic device includes a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus;
[0025] The memory stores a computer program;
[0026] The processor is used to implement the above-mentioned power prediction control method when executing the computer program stored on the memory.
[0027] In a fourth aspect, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned power prediction control method is implemented.
[0028] This disclosure at least includes the following beneficial effects:
[0029] This disclosure analyzes the rolling optimization link based on the principle of generalized predictive control, and analyzes the influence of each parameter in generalized predictive control on the control effect from the principle perspective. This disclosure provides a convenient predictive control method for complex control structures, bringing great convenience to the use in industrial sites. This disclosure provides a mechanistic analysis for the control method and has wide application value.
[0030] Other features and advantages of the present disclosure will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present disclosure. The objectives and other advantages of the present disclosure may be realized and attained by the structure particularly pointed out in the specification and the drawings. Description of the Drawings
[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the following briefly introduces the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0032] Figure 1 Schematic flowchart of the power prediction control method according to an embodiment of the present disclosure;
[0033] Figure 2 Schematic diagram of the control effect of directly adopting generalized predictive control according to an embodiment of the present disclosure;
[0034] Figure 3 Schematic diagram of the control effect when the proportional coefficient is out of balance according to an embodiment of the present disclosure;
[0035] Figure 4 Schematic diagram of the control effect after correcting the proportional coefficient out of balance according to an embodiment of the present disclosure;
[0036] Figure 5 Schematic diagram of the control effect when the inertia time is out of balance according to an embodiment of the present disclosure;
[0037] Figure 6 Schematic diagram of the control effect after correcting the inertia time out of balance according to an embodiment of the present disclosure;
[0038] Figure 7 Schematic diagram of the structure of the power prediction control device according to an embodiment of the present disclosure;
[0039] Figure 8 Schematic diagram of the structure of the electronic device according to an embodiment of the present disclosure. Detailed Embodiments
[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following clearly and completely describes the technical solutions in the embodiments of the present disclosure with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some, but not all, of the embodiments of the present disclosure. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present disclosure without creative efforts fall within the scope of protection of the present disclosure.
[0041] As Figure 1As shown, a power prediction control method, the method comprising:
[0042] S101, establishing a prediction model based on a controlled autoregressive moving average model;
[0043] S102, based on the prediction model, using rolling finite horizon optimization to obtain a local optimal solution and applying it to power control;
[0044] S103, dynamically correcting the error between the prediction model and the actual situation according to the real-time result of power control.
[0045] When specifically implemented, it is introduced as follows:
[0046] The model adopted by generalized predictive control is called a prediction model. Generalized predictive control has relatively low requirements for the model. Compared with traditional control methods, generalized predictive control emphasizes more on the function of the model. That is to say, as long as the model can use the data information from the past to predict the future output behavior of the system, the model can be used as a prediction model. In this disclosure, a controlled autoregressive moving average model is adopted as the prediction model. The autoregressive moving average model is also known as the CARMA model.
[0047] Due to the changes in the production environment and the influence of external disturbances, actual industrial processes are often accompanied by non-linearity, time-variation, and uncertainty. In this case, the so-called optimal control system designed according to the ideal model is only optimal in a mathematical sense, but loses its optimality for actual industrial processes. Moreover, it may even lead to a serious decline in control quality and the inability to produce normally. Generalized predictive control has made great improvements in this regard. The previous optimal control scheme calculates the control scheme for the entire process at one time, and finally realizes the optimality of the entire control process in an off-line calculation manner. In generalized predictive control, the rolling optimization link is used to achieve the local optimality based on the currently known information. The optimization in generalized predictive control does not adopt the off-line calculation of the entire control scheme of the system, but adopts a rolling finite-time domain optimization strategy. In generalized predictive control, the calculation formula of the optimal control scheme is carried out repeatedly online. At each sampling moment, the optimization performance index only focuses on the future finite-time domain. At the current moment, the algorithm considers the control effect of the future finite-time domain and calculates the control scheme for the future finite-time domain, but only sends the control quantity at the current moment to the system. At the next sampling moment, the finite optimization time domain considered by the algorithm is shifted backward, and the above operation is carried out again. Generally speaking, for generalized predictive control, there is an optimization performance index relative to each sampling moment at each sampling moment. The algorithm will calculate the control quantity according to this index and send it to the system. This control scheme of generalized predictive control is a local optimal solution, which is theoretically a sub-optimal solution inferior to the global optimal solution. However, in the industrial field, the environment is very complex, and the controlled object may also change to a certain extent. Since this optimization scheme is carried out repeatedly online and always builds the optimization process on the basis of the latest information obtained from the actual process, the setting of this link can enable the algorithm to correct the uncertainties caused by model mismatch, time-variation, and interference more timely, so it can bring stronger robustness and ensure the actual optimality of the control system.
[0048] Due to the setting of the rolling optimization link, the optimization process can always be based on the latest information obtained from the actual process, and the setting of the feedback correction link further strengthens this point. Due to the existence of uncertain factors such as non-linearity, time-variation, and interference in actual industrial processes, the prediction model used in the prediction algorithm must have a certain deviation from the actual system. This deviation in the model will lead to a decline in the control effect, and in severe cases, it may even lead to the inability to produce normally. In generalized predictive control, a feedback correction method is adopted to compensate for the error between the prediction model and the actual model. In each calculation of rolling optimization, the control algorithm will compensate the output value of the prediction model based on the error between the prediction model and the actual model at the previous moment. After compensation, the output of the prediction model will be close to the output of the actual system to a great extent, thus realizing the feedback correction of the model.
[0049] Generalized Predictive Control (GPC) is a control method based on the system prediction model, and its foundation lies in the identification of the system model. When there is a mismatch between the actual system and the prediction model, the control effect of GPC will also be affected to a certain extent. This disclosure takes the parameters of a second-order transfer function as an example to summarize the impact on GPC when each parameter is mismatched. This disclosure will analyze from two perspectives: overshoot and settling time.
[0050] Generalized Predictive Control is a control based on the system discrete model. The more accurate the model, the better the control effect. When the model is perfectly matched, the parameters that have a gain effect on the control quantity will accelerate the convergence speed of the system (basically, the larger the better). However, after the model mismatch, the control quantity output by the algorithm will, to some extent, worsen the overshoot or settling time of the system. Therefore, when there is a model mismatch, we need to adjust the relevant parameters so that GPC can re - achieve a better control effect for the system.
[0051] In a traditional PID controller, the practical significance of each parameter is relatively obvious, and parameter tuning can be carried out according to the operator's understanding and judgment during industrial field operation. However, in Generalized Predictive Control, the practical physical meaning of each parameter is not obvious. Generally, no parameter tuning is carried out after commissioning, or parameter tuning is completely based on the operator's experience. This parameter - tuning method has relatively high risks. This disclosure uses simulation to mimic the actual parameter - tuning process, and through analyzing the influence of each parameter in GPC on the control quantity, finally sorts out a set of feasible GPC parameter - tuning processes.
[0052] Considering that the controlled object in the industrial field generally adopts an inertia plus delay model, the controlled object adopted in this disclosure is a second - order system composed of two inertia links with a gain of 1 and an inertia time of 170 in series, and a pure delay of 30 seconds. This disclosure takes overshoot and settling time as the evaluation indexes of the control effect, where the smaller the overshoot, the better, and the shorter the settling time, the better. The formulas are as follows:
[0053]
[0054] In the formula, s represents seconds.
[0055] The control effect of directly adopting Generalized Predictive Control is as Figure 2 shown. Since the GPC in this disclosure adopts the CARMA model, the sampling time needs to be set, and the sampling time is set to 1 second.
[0056] The GPC parameters are set as follows:
[0057] Prediction horizon: 150 seconds, control horizon: 8 seconds, softening coefficient: 0.9
[0058] The control effect at this time is:
[0059] Adjustment time: 195 seconds, overshoot: 2.31%.
[0060] When the proportional coefficient is out of balance, for example, changing the gain of the controlled object in the prediction model to 1.5, the GPC control effect is as Figure 3 shown.
[0061] At this time, the adjustment time is 180 seconds and the overshoot is 10.12%
[0062] By comparing the control quantity when the control quantity is adapted to the model, when the model is mismatched, the control quantity may fluctuate greatly in a short time. According to this performance, it is possible to consider increasing the prediction time domain, reducing the control time domain, increasing the softening coefficient, and the adjustment is as shown in Figure 4.
[0063] At this time, the prediction time domain is: 170 seconds, the control time domain is: 5 seconds, and the softening coefficient is 0.95
[0064] At this time, the adjustment time is: 214 seconds and the overshoot is 4.9%
[0065] It can be seen that although part of the adjustment time is sacrificed, the overshoot has been greatly improved.
[0066] When the inertia time is out of balance, for example, changing the inertia time of the controlled object in the prediction model to 200, the GPC control effect is as Figure 5 shown:
[0067] At this time, the adjustment time is 217 seconds and the overshoot is 2.11%
[0068] It can be seen that although the change in the control effect is not obvious, the change range of the control quantity output by GPC is relatively large, and this situation is unacceptable in the actual industrial site. It is possible to consider reducing the control time domain and increasing the softening coefficient, and the adjustment is as Figure 6 shown:
[0069] At this time, the control time domain is: 30 seconds and the softening coefficient is 0.95
[0070] At this time, the adjustment time is: 198 seconds and the overshoot is 1.79%
[0071] It can be seen that while the adjustment time and overshoot are improved, the fluctuation of the control quantity is also suppressed.
[0072] In the part where the prediction time domain is greater than the control time domain, the control quantity does not change, and the control quantity at the end of the control time domain is used as the control quantity in the part where the prediction time domain is greater than the control time domain. Therefore, the influence of the prediction time domain on the algorithm is mainly in terms of the size of the algorithm. By adjusting the size of the prediction time domain, we can directly affect the overshoot and adjustment time of the system. Generally speaking, the smaller the overshoot of the system, the longer the adjustment time, and a balance needs to be made between the two.
[0073] Although the control time domain also affects the change of the control quantity, the principle of the influence of the control time domain on the control quantity is different. The control quantity within the control time domain is calculated according to the prediction model in GPC, and the calculation is performed at each sampling time. When there is a model mismatch, the control quantity calculated within the control time domain is no longer correct. Since the deviation of the model will accumulate, the longer the control time domain, the greater the accumulated deviation, and at this time the control quantity is more incorrect. Therefore, the control time domain should be appropriately reduced. Since the calculation formula of the control quantity includes the Diophantine equation coefficients of the prediction model, the greater the difference between the models, the greater the difference between the Diophantine equation coefficients.
[0074] Considering that the overshoot of the system is relatively large, and the essence of the overshoot can be understood as that the tracking effect of the system is not very good and it cannot accurately track the set value. Therefore, the flexibility coefficient related to the "tracking concept" can be adjusted. The flexibility coefficient can make the set value input into the prediction model smoother, rather than a step change. Reasonably selecting the flexibility coefficient can reduce the oscillation of the system. The larger the flexibility coefficient, the smoother the system can track. The smaller the flexibility coefficient, the faster the system can track, but there is a risk of oscillation. It should be noted that the essence of increasing the flexibility coefficient is also to affect the control effect by reducing the control quantity.
[0075] As Figure 7 shown, a power prediction control device includes: a model establishment unit 701, a control unit 702, and a correction unit 703;
[0076] The model establishment unit 701 is used to establish a prediction model based on the controlled autoregressive moving average model;
[0077] The control unit 702 is used to obtain a local optimal solution based on the prediction model by using rolling finite-time domain optimization and is used for power control;
[0078] The correction unit 703 is used to dynamically correct the error between the prediction model and the actual situation according to the real-time result of power control.
[0079] As Figure 8 shown, the present disclosure provides an electronic device, including a processor 801, a communication interface 802, a memory 803, and a communication bus 804. Among them, the processor 801, the communication interface 802, and the memory 803 complete mutual communication through the communication bus 804;
[0080] The memory 803 stores a computer program;
[0081] The processor 801 is used to implement the above-mentioned power prediction control method when executing the computer program stored on the memory 803.
[0082] The present disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned power prediction control method.
[0083] The computer-readable storage medium may be included in the device / apparatus described in the above embodiments; or it may exist alone without being assembled into the device / apparatus. The above computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present disclosure.
[0084] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example, but not limited to: portable computer disks, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device, or device.
[0085] Although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.
Claims
1. A power prediction control method, characterized in that: The method comprises: Establish a forecasting model based on the controlled autoregressive moving average model; Based on the prediction model, a rolling finite horizon optimization is used to obtain the local optimal solution and use it for power control; According to the real-time results of power control, the error between the prediction model and the actual situation is dynamically corrected.
2. A power prediction control method according to claim 1, characterized in that: The local optimal solution is obtained by rolling finite horizon optimization and used for power control, including: At the current moment, based on the control effect of the finite time domain, a control scheme of the finite time domain is calculated; Obtain the local optimal solution from the control scheme and send the control quantity at the current moment to the power system; At the next sampling moment, the finite time domain moves backward and the local optimal solution is recalculated.
3. The power prediction control method according to claim 1, characterized in that: The prediction model adopts a second-order system consisting of two inertia links in series with preset gain and inertia time of the control object.
4. The power prediction control method according to claim 3, characterized in that: Dynamically correct the error between the prediction model and the actual situation, including: The error between the prediction model and the actual situation is determined by the overshoot and adjustment time.
5. The power prediction control method according to claim 3, characterized in that: Dynamically correct the error between the prediction model and the actual situation, including: The error between the prediction model and the actual situation is dynamically corrected through the prediction time domain, control time domain and softening coefficient.
6. A power prediction control method according to claim 5, characterized in that: When the proportional coefficient of the prediction model is out of adjustment, the prediction time domain is increased, the control time domain is reduced, and the softening coefficient is increased.
7. The power prediction control method according to claim 5, characterized in that: When the inertia time of the prediction model is out of adjustment, the control time domain is reduced and the softening coefficient is increased.
8. A power prediction control device, characterized in that: include: A model building unit, a control unit and a correction unit; A model building unit, used for building a prediction model based on a controlled autoregressive moving average model; A control unit is used to obtain a local optimal solution based on a prediction model and a rolling finite horizon optimization, and is used for power control; The correction unit is used to dynamically correct the error between the prediction model and the actual situation according to the real-time results of power control.
9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; a memory storing a computer program; The processor is used to implement a power prediction control method according to any one of claims 1 to 7 when executing a computer program stored in a memory.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, an electric power prediction control method according to any one of claims 1 to 7 is implemented.