Grid-connected unit maintenance optimization method, system, equipment and medium

Abstract

The invention provides a grid-connected unit maintenance optimization method, system and device and a medium, and relates to the technical field of power systems. Comprising the following steps: carrying out load prediction error analysis on new energy aiming at the uncertainty influence of a system to obtain a load prediction value; according to the load prediction value, a fuzzy set based on Wasserstein measurement is constructed; taking comprehensive economy as an optimization target, and adopting a fuzzy set to establish a unit maintenance model with multiple constraint conditions; the unit maintenance model adopts distribution robust opportunity constraint, and a maintenance strategy arrangement problem is linearized and reconstructed into a mixed integer linear programming problem; according to the method, the optimal maintenance plan is obtained under the uncertain condition, the economical efficiency is improved, the reliability and the safety of the system are comprehensively considered, and the method is suitable for the scene of high-proportion new energy grid connection in a modern electric power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, system, equipment and medium for optimizing maintenance of grid-connected units. Background Art

[0002] In response to the problem of high power abandonment caused by the access of new energy to the power system, most of the research focuses on aspects such as unit combination and energy storage configuration, and improves the new energy consumption rate by adjusting the startup sequence of the units and configuring a certain amount of energy storage. In addition, the maintenance of units in the system will also have many impacts on the consumption of new energy. The formulation of the maintenance plan for generator sets is an important part of the planning and operation of the power system, and the maintenance and shutdown of generator sets involves many problems. If the maintenance plan is not arranged properly, it will have many adverse effects on the reliability and economy of the power system.

[0003] At present, in the arrangement of unit maintenance plans, the power grids in various regions still mainly use the equal reserve method, that is, by arranging unit maintenance, the system's reserve capacity or reserve rate is roughly equal in each period of the year. However, with the increase in the penetration rate of new energy, the volatility of new energy output has an important impact on the operation of the system. If the equal reserve method is continued to be used to formulate unit maintenance plans for high-proportion new energy power systems, serious new energy power abandonment problems may occur. Therefore, the use of distributed robust uncertainty sets to construct wind power uncertainty models can take into account calculation time, economy and robustness at the same time, and has wide applicability and research value. Summary of the invention

[0004] The main purpose of the present invention is to provide a grid-connected unit maintenance optimization method, system, equipment and medium to solve the technical problems of long calculation time, low economy, insufficient robustness and low applicability in current unit maintenance.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a maintenance optimization method for grid-connected units, comprising the following steps: Step 1: In view of the uncertainty of the system, according to the load characteristics of new energy and combined with historical load data, the LSTM model is used for prediction, prediction data is generated, and load prediction error analysis of new energy is performed to obtain load prediction values; Step 2: According to the load forecast value, a fuzzy set based on Wasserstein metric is constructed to transform the uncertainty of the new energy load into a processable mathematical model, and the membership value of the fuzzy set is normalized; Step 3: Taking comprehensive economy as the optimization target, the fuzzy set in step S2 is used to establish a unit maintenance model with multiple constraints; Step 4: The unit maintenance model uses distributed robust chance constraints to linearize the maintenance strategy scheduling problem and reconstruct it into a mixed integer linear programming problem; Step 5: Construct the objective function in the unit maintenance model based on system reliability, system safety and renewable energy power abandonment, and then solve it to obtain the optimal unit maintenance result.

[0006] In the preferred solution, in step 5, the objective function adopts maximizing system reliability, minimizing the amount of new energy abandoned, and maximizing system safety. Among them, maximizing system reliability converts the system reliability into the cost of unit startup and power generation fuel related to the use of spare capacity during maintenance. The formula is: ; In the formula, The unit at time t i The fuel cost, Start-up cost for standby units; T Gather for the moment; R G Assemble for the crew; Minimizing the amount of abandoned renewable energy is to convert the abandoned amount into economic cost and calculate the economic loss caused by the abandoned electricity. The formula is: ; In the formula, for t The amount of renewable energy wasted at any given moment (in MWh), for t The economic cost of the amount of renewable energy wasted per unit of time; Maximizing system safety converts system safety into calculated economic losses or costs caused by equipment failures, using the formula: ; In the formula, I is the set of all accident types, It is an event i The weight of reflects its impact on the overall system security. It is an event i exist t The probability of occurrence at any moment, It is an event i exist t The economic losses caused by the moment; After unifying the dimensions of the three, the final economic cost objective function is formed: .

[0007] In the preferred solution, in step 1, the load forecast error analysis is performed on the new energy source, and the load forecast error formula is: ; In the formula, , , The node loads are j In the period t prediction error, actual power and predicted power; The short-term load forecast error usually follows a normal distribution with a mean of 0, and the standard deviation formula is: ; k It is a certain proportional value, which is determined according to the specific load conditions.

[0008] In the preferred solution, in step 2, the fuzzy set of Wasserstein metric is used to reflect the uncertainty of the load forecast value, including: S201: Construct empirical distribution based on load forecast value , the formula is: ; Where: For sample N The empirical distribution of the composition; For The Dirac distribution centered at ; N is the number of historical samples; S202: Combined with the Wasserstein metric, we get The distribution of is: ; Where: W(X , X N ) is the Wasserstein metric; , They are respectively distributed X , X N A random variable of S203: Fuzzy sets based on Wasserstein metric D w Can be constructed as: ; Where: D w is a fuzzy set based on Wasserstein metric; is the Wasserstein metric radius; M is the set of all distributions.

[0009] In the preferred solution, in step 3, a unit maintenance model with multiple constraints is established, including: 301: Considering that the power on the load side is constantly changing, the load power constraint is replaced by the load power opportunity constraint. The formula is: ; Where: and Load resources j At the moment t Maximum and minimum power; is the confidence level; S302: After optimizing the load power balance constraint into a chance constraint, reconstruction is performed using distributed robust chance constraint optimization, and the formula is: ; Where: v , z m , q m is an auxiliary variable, M is a larger constant, is the Wasserstein metric radius.

[0010] A grid-connected unit maintenance optimization system, comprising: The load error module is used to target the uncertainty impact of the system, based on the load characteristics of new energy, combined with historical load data, select the LSTM model for prediction, generate prediction data and perform load prediction error analysis on new energy to obtain load prediction values; The fuzzy set module is used to construct a fuzzy set based on Wasserstein metric according to the load forecast value, transform the uncertainty of the new energy load into a processable mathematical model, and normalize the membership value of the fuzzy set; The maintenance model module is used to establish a unit maintenance model with multiple constraints by using the fuzzy set in step S2 with comprehensive economic type as the optimization target; The opportunity constraint module is used in the unit maintenance model to linearize the maintenance strategy scheduling problem and reconstruct it into a mixed integer linear programming problem by using distributed robust opportunity constraints; The unit optimization module is used to construct the objective function in the unit maintenance model based on system reliability, system safety and renewable energy power abandonment, and then solve it to obtain the optimal unit maintenance result.

[0011] An electronic device comprises: a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the computer program implements the maintenance optimization method of a grid-connected unit when executed by the processor.

[0012] A computer-readable storage medium comprises: a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for optimizing the maintenance of a grid-connected unit is implemented.

[0013] The present invention provides a maintenance optimization method for grid-connected units. According to the load characteristics of new energy, load forecasting errors of new energy are analyzed to obtain load forecast values. The Wasserstein metric is used to construct fuzzy sets to effectively capture the difference between the forecast value and the actual value. The method is suitable for describing uncertainty and determining various constraints in the maintenance model. The maintenance strategy scheduling problem is converted into a form that is easy to solve by distributing Robust Chance Constraints. The complex nonlinear problem is linearized and reconstructed into a mixed integer linear programming. The objective function is optimized on the basis of considering system reliability, safety and the amount of abandoned power of new energy, and the best unit maintenance strategy is finally obtained. This embodiment introduces Distributed Robust Chance Constraints and Wasserstein Metrics to convert the complex grid-connected unit maintenance optimization problem into a solvable mixed integer linear programming problem, thereby obtaining the optimal maintenance plan under uncertainty conditions, which not only improves the economy, but also comprehensively considers the reliability and safety of the system, and is suitable for scenarios where a high proportion of new energy is connected to the grid in modern power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of the maintenance optimization method of the present invention; Figure 2 is a flow chart of new energy load prediction of the present invention; Figure 3 is a cost comparison chart of different detection schemes of the present invention; Figure 4 It is a diagram of the internal structure of the computer device of the present invention. DETAILED DESCRIPTION

[0015] Example 1 like Figure 1 As shown, a method for optimizing the maintenance of a grid-connected unit comprises the following steps: Step 1: In view of the uncertainty of the system, according to the load characteristics of new energy and combined with historical load data, the LSTM model is selected for prediction, prediction data is generated, and load prediction error analysis of new energy is performed to obtain the load prediction value.

[0016] Step 2: According to the load forecast value, a fuzzy set based on Wasserstein metric is constructed to transform the uncertainty of the new energy load into a processable mathematical model, and the membership value of the fuzzy set is normalized.

[0017] Step 3: Taking comprehensive economy as the optimization target, the fuzzy set in step S2 is used to establish a unit maintenance model with multiple constraints.

[0018] Step 4: The unit maintenance model uses distributed robust chance constraints to linearize the maintenance strategy scheduling problem and reconstruct it into a mixed integer linear programming problem.

[0019] Step 5: Construct the objective function in the unit maintenance model based on system reliability, system safety and renewable energy power abandonment, and then solve it to obtain the optimal unit maintenance result.

[0020] In this embodiment, according to the load characteristics of new energy, the load forecast error analysis of new energy is carried out to obtain the load forecast value, and the Wasserstein metric is used to construct a fuzzy set to effectively capture the difference between the predicted value and the actual value, which is suitable for the description of uncertainty, and various constraints in the maintenance model are determined. Through the distributed robust opportunity constraint, the maintenance strategy scheduling problem is converted into a form that is easy to solve, and the complex nonlinear problem is linearized and reconstructed into a mixed integer linear programming. On the basis of considering the system reliability, safety and the amount of new energy abandoned electricity, the objective function is optimized, and finally the best unit maintenance strategy is obtained. This embodiment introduces the distributed robust opportunity constraint and the Wasserstein metric to convert the complex grid-connected unit maintenance optimization problem into a solvable mixed integer linear programming problem, so as to obtain the optimal maintenance plan under uncertainty conditions, which not only improves the economy, but also comprehensively considers the reliability and safety of the system, and is suitable for the scenario of a high proportion of new energy grid-connected in modern power systems.

[0021] The membership values ​​of the fuzzy sets are normalized to allow for consistent comparison for subsequent model constraint optimization.

[0022] In the preferred solution, in step 5, the objective function adopts maximizing system reliability, minimizing the amount of new energy abandoned, and maximizing system safety. Among them, maximizing system reliability converts the system reliability into the cost of unit startup and power generation fuel related to the use of spare capacity during maintenance. The formula is: ; In the formula, The unit at time t i The fuel cost, Start-up cost for standby units; T Gather for the moment; R G During the overhaul process, the reliability of the system should be kept as high as possible.

[0023] The goal of minimizing the amount of abandoned renewable energy is to reduce the waste of renewable energy power generation due to insufficient system load or other problems. It is to convert the abandoned amount into economic cost and calculate the economic loss caused by abandoned power. The formula is: ; In the formula, for t The amount of renewable energy wasted at any given moment (in MWh), for t The economic cost of the amount of renewable energy wasted per unit time.

[0024] By optimizing the unit maintenance plan, we can ensure that more new energy can be effectively utilized.

[0025] Maximizing system safety converts system safety into calculated economic losses or costs caused by equipment failures, using the formula: ; In the formula, I is the set of all accident types, It is an event i The weight of reflects its impact on the overall system security. It is an event i exist t The probability of occurrence at any moment, It is an event i exist t The economic losses caused when the moment occurs.

[0026] Maximizing system safety ensures safe operation of the system during maintenance and avoids system instability or safety hazards caused by maintenance.

[0027] After unifying the dimensions of the three, the final economic cost objective function is formed: .

[0028] By relating system reliability to reserve capacity and fuel costs, the model can more realistically consider the economics and practicality of maintenance scheduling.

[0029] In the preferred solution, in step 1, in view of the uncertainty impact of the system, according to the load characteristics of new energy and combined with historical load data, the LSTM model is selected for prediction.

[0030] like Figure 2As shown in the figure, first input the collected new energy load data, and then slide and update the input data, and reasonably set the value and order of the input influencing factors. Then, the input of the model is weighted through the attention mechanism, the effective features are highlighted and a new input vector is formed, and then the vector is input into the bidirectional LSTM network for training; finally, the output after network training is input into the linear transformation layer and the Softmax layer to predict the new energy load results.

[0031] Based on the load forecasting results of the model, the load forecasting error analysis of new energy is carried out, and the load forecasting error formula is: ; In the formula, , , The node loads are j In the period t prediction error, actual power and predicted power; The short-term load forecast error usually follows a normal distribution with a mean of 0, and the standard deviation formula is: ; k It is a certain proportional value, which is determined according to the specific load conditions.

[0032] Through load forecasting error analysis, the error rate is reduced and the optimization accuracy is improved. k It can be changed according to the empirical value, thus affecting the robustness analysis and response strategy of the system in the case of load forecast error.

[0033] In the preferred solution, in step 2, the fuzzy set of Wasserstein metric is used to reflect the uncertainty of the load forecast value, including: S201: Construct empirical distribution based on load forecast value , the formula is: ; Where: For sample N The empirical distribution of the composition; For The Dirac distribution centered at ; N is the number of historical samples.

[0034] S202: Combined with the Wasserstein metric, we get The distribution of is: ; Where: W(X , X N )is the Wasserstein metric; , They are respectively distributed X , X N A random variable.

[0035] S203: Fuzzy sets based on Wasserstein metric D w Can be constructed as: ; Where: D w is a fuzzy set based on Wasserstein metric; is the Wasserstein metric radius; M is the set of all distributions.

[0036] According to the prediction error in step 1, the Wasserstein metric is used to construct fuzzy sets. The Wasserstein metric is used to measure the distance between fuzzy sets to more accurately capture and express the fuzziness and uncertainty of load forecast values.

[0037] Furthermore, fuzzy sets are used to represent the uncertainty of load forecast values ​​by defining fuzzy membership functions. For example, the forecast value and its possible fluctuation range can be represented in the form of fuzzy numbers or fuzzy sets.

[0038] Furthermore, the constructed fuzzy sets are integrated into the subsequent unit maintenance model. When optimizing the unit maintenance strategy, the uncertainty of load forecasting can be taken into account and a better maintenance plan can be formulated.

[0039] Finally, the optimal maintenance strategy is obtained by optimizing and solving the unit maintenance model, taking into account the uncertainty represented by fuzzy sets, so as to have better practical application effect when formulating maintenance plans.

[0040] Through the above steps, the uncertainty of load forecast values ​​can be quantified as fuzzy sets under the Wasserstein metric, which can better reflect the actual fluctuation and uncertainty of load forecast values ​​and help deal with the uncertainty of load forecasting in the optimization and decision-making process.

[0041] In the preferred solution, in step 3, a unit maintenance model with multiple constraints is established, including: 301: Considering that the power on the load side is constantly changing, the load power constraint is replaced by the load power opportunity constraint. The formula is: ; Where: and Load resourcesj At the moment t Maximum and minimum power; is the confidence level.

[0042] S302: After optimizing the load power balance constraint into a chance constraint, reconstruction is performed using distributed robust chance constraint optimization, and the formula is: ; Where: v , z m , q m is an auxiliary variable, M is a larger constant, is the Wasserstein metric radius.

[0043] Distributed Robust Chance Constrained Optimization,In the optimization process, the chance constraint of the load power balance constraint can be further strengthened by distributed robust chance constraint optimization, taking into account the uncertainty of the distribution, and constraining the range of distribution changes through the Wasserstein metric. Combining chance constraints and distributed robust optimization makes the model more robust in the face of distribution uncertainty.

[0044] By replacing the load power constraint with a chance constraint and introducing distributionally robust optimization, the unit maintenance model can better cope with the uncertainty of load power. At the same time, the Wasserstein metric is used to measure the differences between different distributions during the optimization process, which enhances the robustness of the model and ensures that the optimization results of the model are still highly feasible and reliable at a high confidence level.

[0045] This embodiment mainly studies the power grid in a certain region of my country in 2023 under the condition of a high proportion of new energy access. Through the traditional periodic maintenance method and the maintenance optimization method proposed in the present invention, the new energy power generation and load forecast curves are used to simulate and calculate the relevant data of unit maintenance costs under the two maintenance strategies through MATLAB2021a software, and conduct comparative analysis.

[0046] like Figure 3 As shown in the figure, there are obvious differences in the maintenance results under different maintenance strategies. Compared with the current regular maintenance method, the optimized method has obvious advantages in reducing the maintenance cost of new energy.

[0047] Example 2 Further described in conjunction with Example 1, a grid-connected unit maintenance optimization system includes: The load error module is used to target the uncertainty impact of the system, based on the load characteristics of new energy, combined with historical load data, select the LSTM model for prediction, generate prediction data and perform load prediction error analysis on new energy to obtain load prediction values; The fuzzy set module is used to construct a fuzzy set based on Wasserstein metric according to the load forecast value, transform the uncertainty of the new energy load into a processable mathematical model, and normalize the membership value of the fuzzy set; The maintenance model module takes comprehensive economy as the optimization target and uses the fuzzy set in step S2 to establish a unit maintenance model with multiple constraints; The opportunity constraint module is used in the unit maintenance model to linearize the maintenance strategy scheduling problem and reconstruct it into a mixed integer linear programming problem by using distributed robust opportunity constraints; The unit optimization module is used to construct the objective function in the unit maintenance model based on system reliability, system safety and renewable energy power abandonment, and then solve it to obtain the optimal unit maintenance result.

[0048] like Figure 4 As shown, an electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, a grid-connected unit maintenance optimization method based on distributed robust opportunity constraints in Example 1 is implemented.

[0049] A computer-readable storage medium includes: a computer program stored on the computer-readable storage medium, and when the computer program is executed by a processor, a grid-connected unit maintenance optimization method based on distributed robust opportunity constraints in Example 1 is implemented.

[0050] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A maintenance optimization method for grid-connected units, characterized in that: The following steps are involved: Step 1: In view of the uncertainty of the system, according to the load characteristics of new energy and combined with historical load data, the LSTM model is used for prediction, prediction data is generated, and load prediction error analysis of new energy is performed to obtain load prediction values; Step 2: According to the load forecast value, a fuzzy set based on Wasserstein metric is constructed to transform the uncertainty of the new energy load into a processable mathematical model, and the membership value of the fuzzy set is normalized; Step 3: Taking comprehensive economy as the optimization target, the fuzzy set in step S2 is used to establish a unit maintenance model with multiple constraints; Step 4: The unit maintenance model uses distributed robust chance constraints to linearize the maintenance strategy scheduling problem and reconstruct it into a mixed integer linear programming problem; Step 5: Construct the objective function in the unit maintenance model based on system reliability, system safety and renewable energy power abandonment, and then solve it to obtain the optimal unit maintenance result.

2. The grid-connected unit maintenance optimization method according to claim 1, characterized in that: In step 5, the objective function is to maximize system reliability, minimize the amount of new energy curtailment, and maximize system safety. Maximizing system reliability converts the system reliability into the cost of unit startup and power generation fuel related to the use of spare capacity during maintenance. The formula is: ； In the formula, The unit at time t i The fuel cost, Start-up cost for standby units; T Gather for the moment; R G Assemble for the crew; Minimizing the amount of abandoned renewable energy is to convert the abandoned amount into economic cost and calculate the economic loss caused by the abandoned electricity. The formula is: ； In the formula, for t The amount of renewable energy wasted at any given moment (in MWh), for t The economic cost of the amount of renewable energy wasted per unit of time; Maximizing system safety converts system safety into calculated economic losses or costs caused by equipment failures, using the formula: ； In the formula, I is the set of all accident types, It is an event i The weight of reflects its impact on the overall system security. It is an event i exist t The probability of occurrence at any moment, It is an event i exist t The economic losses caused by the moment; After unifying the dimensions of the three, the final economic cost objective function is formed: 。 3. The grid-connected unit maintenance optimization method according to claim 1, characterized in that: In step 1, the load forecast error analysis of new energy is carried out, and the load forecast error formula is: ； In the formula, , , The node loads are j In the period t prediction error, actual power and predicted power; The short-term load forecast error usually follows a normal distribution with a mean of 0, and the standard deviation formula is: ； k It is a certain proportional value, which is determined according to the specific load conditions.

4. The grid-connected unit maintenance optimization method according to claim 1, characterized in that: In step 2, the fuzzy set of Wasserstein metric is used to reflect the uncertainty of load forecast value, including: S201: Construct empirical distribution based on load forecast value , the formula is: ； Where: For sample N The empirical distribution of the composition; For The Dirac distribution centered at ; N is the number of historical samples; S202: Combined with the Wasserstein metric, we get The distribution of is: ； Where: W(X , X N ) is the Wasserstein metric; , They are respectively distributed X , X N A random variable of S203: Fuzzy sets based on Wasserstein metric D w Can be constructed as: ； Where: D w is a fuzzy set based on Wasserstein metric; is the Wasserstein metric radius; M is the set of all distributions.

5. The grid-connected unit maintenance optimization method according to claim 1, characterized in that: In step 3, a unit maintenance model with multiple constraints is established, including: 301: Considering that the power on the load side is constantly changing, the load power constraint is replaced by the load power opportunity constraint. The formula is: ； Where: and Load resources j At the moment t Maximum and minimum power; is the confidence level; S302: After optimizing the load power balance constraint into a chance constraint, reconstruction is performed using distributed robust chance constraint optimization, and the formula is: ； Where: v , z m , q m is an auxiliary variable, M is a larger constant, is the Wasserstein metric radius.

6. A grid-connected unit maintenance optimization system, characterized in that: include: The load error module, in view of the uncertainty of the system, uses the LSTM model for prediction based on the load characteristics of new energy sources and historical load data, generates prediction data, and performs load prediction error analysis on new energy sources to obtain load prediction values; The fuzzy set module is used to construct a fuzzy set based on Wasserstein metric according to the load forecast value, transform the uncertainty of the new energy load into a processable mathematical model, and normalize the membership value of the fuzzy set; The maintenance model module is used to establish a unit maintenance model with multiple constraints by using the fuzzy set in step S2 with comprehensive economic type as the optimization target; The opportunity constraint module is used in the unit maintenance model to linearize the maintenance strategy scheduling problem and reconstruct it into a mixed integer linear programming problem by using distributed robust opportunity constraints; The unit optimization module is used to construct the objective function in the unit maintenance model based on system reliability, system safety and renewable energy power abandonment, and then solve it to obtain the optimal unit maintenance result.

7. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the grid-connected unit maintenance optimization method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: include: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the grid-connected unit maintenance optimization method according to any one of claims 1 to 5 is implemented.

Images