Multi-card multi-model scheduling method based on greedy method

Abstract

The invention belongs to the technical field of power dispatching, and relates to a multi-card multi-model dispatching method based on a greedy method, which performs prediction based on key variables such as load, generating capacity, equipment state, energy storage state and energy price, and generates a preliminary dispatching strategy by using a heuristic algorithm based on the prediction results. A greedy algorithm is used for optimizing a preliminary scheduling strategy, and a global optimal solution is continuously approached through local optimal selection of the greedy algorithm, so that a relatively good scheduling effect can still be realized under the condition that a prediction error exists. Finally, the scheduling strategy is evaluated by comparing the difference between the actual operation data and the prediction result in real time, parameters of the greedy algorithm are dynamically adjusted according to the evaluation result, and the defects of the preliminary strategy are corrected. The closed-loop optimization process ensures that the scheduling strategy can adaptively deal with the prediction error and the uncertainty of the system, the scheduling accuracy and efficiency are improved, and the core technical problem of how to optimize the power grid scheduling strategy in the uncertainty is effectively solved.

Description

Technical Field

[0001] The present application belongs to the technical field of power dispatching, and more specifically, to a multi-card multi-model dispatching method based on a greedy method. Background Art

[0002] With the transformation of the global energy structure and the rapid growth of the proportion of renewable energy power generation, modern power systems are facing increasingly complex operational challenges. Traditional power grids mainly rely on centralized power generation such as thermal power, and dispatching is highly controllable and predictable. However, with the access of new energy sources such as wind power and photovoltaic power, which are intermittent, random, and uncertain, the complexity of power grid dispatching has increased significantly. At the same time, in order to improve energy utilization efficiency, new equipment and technologies such as energy storage systems, electric vehicles, and distributed power generation have gradually been integrated into the power system, making it necessary for power grid dispatching to not only consider real-time supply and demand balance, but also to take into account economy, environmental impact, and equipment safety.

[0003] Multi-model prediction technology has become one of the key technologies in modern power grid dispatching. By independently modeling and predicting different variables (such as load, power generation, energy storage status, equipment status, energy prices, etc.), the multi-model prediction method can provide refined and accurate prediction results to support power grid dispatching decisions.

[0004] Although multi-model prediction provides higher accuracy and flexibility for power grid dispatching, how to generate optimized dispatching strategies under the premise of errors and uncertainties in multi-model prediction results, and dynamically adjust the optimization parameters through real-time feedback mechanism to improve the accuracy and efficiency of dispatching strategies are still technical problems that need to be solved. Summary of the invention

[0005] The present invention provides a multi-card multi-model scheduling method based on the greedy method, which aims to solve the technical problem of how to generate an optimized scheduling strategy under the premise that there are errors and uncertainties in the multi-model prediction results, and dynamically adjust the optimization parameters through a real-time feedback mechanism.

[0006] The multi-card multi-model scheduling method based on the greedy method includes the following steps:

[0007] Step 1: Collect real-time monitoring data of the power grid, historical operation data, meteorological data, equipment parameters and energy prices, and pre-process the collected data;

[0008] Step 2: Based on the preprocessed data, multiple models are used to perform predictions to obtain the prediction results of load, power generation, equipment status, energy storage status and energy price respectively;

[0009] Step 3: Based on the prediction results, a heuristic algorithm is used to generate a preliminary scheduling strategy;

[0010] Step 4: Based on the generated preliminary scheduling strategy, a greedy algorithm is used to optimize the preliminary scheduling strategy to obtain an optimized scheduling strategy;

[0011] Step 5: Compare the actual operation data with the model prediction results to evaluate the performance of the scheduling strategy. According to the evaluation results, adjust the parameters of the greedy algorithm to improve the accuracy and efficiency of the scheduling strategy.

[0012] The multi-card multi-model scheduling method based on the greedy method of the present invention effectively solves the technical problem of how to generate an optimized scheduling strategy and dynamically adjust the optimization parameters through a real-time feedback mechanism under the premise that there are errors and uncertainties in the prediction results of multiple models. First, by collecting and preprocessing the real-time monitoring data, historical data, meteorological data, etc. of the power grid, accurate input is provided for multi-model prediction, thereby reducing the source of prediction errors. Secondly, multiple prediction models are used to predict key variables such as load, power generation, equipment status, energy storage status and energy price to improve the accuracy of the prediction results, comprehensively consider the advantages and disadvantages of different models, and reduce the deviation that may be caused by a single model. Next, based on these prediction results, a preliminary scheduling strategy is generated using a heuristic algorithm to ensure the rationality and feasibility of the scheduling decision. On this basis, the greedy algorithm is used to optimize the preliminary scheduling strategy, and the global optimal solution is continuously approached through the local optimal selection of the greedy algorithm, so that a better scheduling effect can be achieved in the presence of prediction errors. Finally, the scheduling strategy is evaluated by comparing the difference between the actual operation data and the prediction results in real time, and the parameters of the greedy algorithm are dynamically adjusted according to the evaluation results to correct the shortcomings of the preliminary strategy. This closed-loop optimization process ensures that the dispatch strategy can adaptively respond to prediction errors and system uncertainties, improves the accuracy and efficiency of dispatch, and effectively solves the core technical problem of how to optimize power grid dispatch strategies under uncertainty.

[0013] Preferably, the objective function and constraints of the heuristic algorithm in step 3 are as follows:

[0014] Objective function:

[0015]

[0016] Where: C t represents the power generation in the tth time period; G t represents the power generation cost in the tth time period; D t represents the energy storage discharge cost in the tth time period; S t represents the energy storage discharge in the tth time period; P t represents energy price; B t represents the amount of electricity purchased in the tth time period; Q t represents the electricity price in the tth time period; T represents the dispatching period;

[0017] Constraints:

[0018]

[0019] G min,i ≤G i (t)≤G max,i ;

[0020] E min,i ≤E i (t)≤E max,i ;

[0021] P t ≥0;

[0022] Q t ≥0;

[0023] Where: G i (t) represents the output of power generation unit i in time period t; L t Represents the load demand forecast result; G min,i represents the minimum output of power generation unit i; G max,i Indicates the maximum output of the power generation unit; E min,i represents the minimum state limit of energy storage system i; E max,i Indicates the maximum state limit of the energy storage system; E i (t) represents the state of energy storage system i in time period t.

[0024] Preferably, step 3 comprises the following steps:

[0025] Initialization: Based on the forecast results of load, power generation, equipment status, energy storage status and energy price, initialize the status of each power generation unit and energy storage system, and set the processing of each power generation unit and the charging and discharging status of the energy storage system to the minimum value that can be operated at present;

[0026] Objective function initialization: Initialize the objective function F to evaluate the pros and cons of the scheduling strategy;

[0027] Load distribution: According to the load demand prediction result L t And the output limit of the power generation unit, allocate the power generation task:

[0028] For each power generation unit i, calculate its maximum output G max,i , distribute the load according to cost efficiency, give priority to the lowest cost power generation unit, and update G in the objective function F t ;

[0029] Energy storage dispatch: E is predicted based on energy storage status t and energy price forecasts t Energy storage scheduling:

[0030] If the energy storage cost is lower than the market electricity purchase price, energy storage discharge is used first; if the market electricity price is lower than the energy storage discharge cost, energy storage charging is performed; update S in the objective function F t and B t ;

[0031] Balancing adjustment: Using constraints to adjust the processing of power generation units and energy storage systems to ensure that the constraints are met;

[0032] Objective function optimization: By adjusting G t , S t and B t To minimize the objective function F, repeat the steps from load distribution to balance adjustment until the objective function F reaches the minimum value or satisfies the predetermined convergence condition;

[0033] Generate preliminary scheduling strategy: According to the optimized G t , S t and B t , generate a preliminary scheduling strategy.

[0034] Preferably, step 4 comprises the following steps:

[0035] Initialization and data input: According to the preliminary scheduling strategy, the output of the power generation unit, the energy storage status, the energy storage charge and discharge amount, and the objective function are obtained;

[0036] Load distribution of power generation units: Based on the current load demand and the unit cost of the power generation unit, the load is redistributed in accordance with the principle of minimizing the power generation cost:

[0037]

[0038] Where: represents unit fuel cost; G max,i represents the maximum output of power generation unit i; Indicates unit operation and maintenance cost;

[0039] The output of each power generation unit i According to the load demand prediction result L t and unit cost C i Adjustments are made to select the lowest cost generating unit for load distribution:

[0040]

[0041] Always: represents the power generation unit with the minimum cost in the current time period;

[0042] Energy storage system scheduling: The energy storage system scheduling strategy is based on the energy price P t and the energy storage system state S tThe prediction results are used to optimize:

[0043] If the energy storage is discharged, that is, B t >0, the discharge amount is as follows:

[0044]

[0045] Where: E t Indicates the current state of the energy storage system, that is, the current stored electrical energy; Represents the discharge efficiency of the energy storage system; C max Indicates the maximum discharge power of the energy storage system;

[0046] If the energy storage is charged, that is, B t <0, the charge capacity is as follows:

[0047]

[0048] Where: Represents the charging efficiency of the energy storage system; C max Indicates the maximum charging power of the energy storage system; E max Indicates the maximum capacity of the energy storage system; S t Indicates the current energy storage status of the energy storage system, that is, the stored electrical energy;

[0049] Balance adjustment: According to load demand and equipment status, adjust the scheduling of power generation and energy storage systems to minimize the objective function F:

[0050]

[0051] Adjust by constraints and B t Guarantee the following conditions:

[0052] The total amount of power generation matches the load demand:

[0053]

[0054] The energy storage system charges and discharges to meet power limits:

[0055] -C max ≤B t ≤C max ;

[0056] Where: C max Indicates the maximum charging or discharging power limit of the energy storage system;

[0057] The energy storage state must not exceed the capacity limit:

[0058] 0≤S t ≤E max ;

[0059] Where: E max Indicates the maximum capacity of the energy storage system;

[0060] Iteration termination: When the change of the objective function F is less than the preset threshold, the algorithm is terminated and the optimized scheduling strategy is output.

[0061] The beneficial effects of the present invention include:

[0062] The multi-card multi-model scheduling method based on the greedy method of the present invention effectively solves the technical problem of how to generate an optimized scheduling strategy and dynamically adjust the optimization parameters through a real-time feedback mechanism under the premise that there are errors and uncertainties in the prediction results of multiple models. First, by collecting and preprocessing the real-time monitoring data, historical data, meteorological data, etc. of the power grid, accurate input is provided for multi-model prediction, thereby reducing the source of prediction errors. Secondly, multiple prediction models are used to predict key variables such as load, power generation, equipment status, energy storage status and energy price to improve the accuracy of the prediction results, comprehensively consider the advantages and disadvantages of different models, and reduce the deviation that may be caused by a single model. Next, based on these prediction results, a preliminary scheduling strategy is generated using a heuristic algorithm to ensure the rationality and feasibility of the scheduling decision. On this basis, the greedy algorithm is used to optimize the preliminary scheduling strategy, and the global optimal solution is continuously approached through the local optimal selection of the greedy algorithm, so that a better scheduling effect can be achieved in the presence of prediction errors. Finally, the scheduling strategy is evaluated by comparing the difference between the actual operation data and the prediction results in real time, and the parameters of the greedy algorithm are dynamically adjusted according to the evaluation results to correct the shortcomings of the preliminary strategy. This closed-loop optimization process ensures that the dispatch strategy can adaptively respond to prediction errors and system uncertainties, improves the accuracy and efficiency of dispatch, and effectively solves the core technical problem of how to optimize power grid dispatch strategies under uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0064] Figure 1 An overall step block diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0066] See also Figure 1 As shown, the best embodiment of the present invention is further described;

[0067] The multi-card multi-model scheduling method based on the greedy method includes the following steps:

[0068] Step 1: Collect real-time monitoring data of the power grid, historical operation data, meteorological data, equipment parameters and energy prices, and pre-process the collected data;

[0069] Step 2: Based on the preprocessed data, multiple models are used to perform predictions to obtain the prediction results of load, power generation, equipment status, energy storage status and energy price respectively;

[0070] Step 3: Based on the prediction results, a heuristic algorithm is used to generate a preliminary scheduling strategy;

[0071] The objective function and constraints of the heuristic algorithm in step 3 are as follows:

[0072] Objective function:

[0073]

[0074] Where: C t represents the power generation in the tth time period; G t represents the power generation cost in the tth time period; D t represents the energy storage discharge cost in the tth time period; S t represents the energy storage discharge in the tth time period; P t represents energy price; B t represents the amount of electricity purchased in the tth time period; Q t represents the electricity price in the tth time period; T represents the dispatching period;

[0075] Constraints:

[0076]

[0077] G min,i ≤G i (t)≤G max,i ;

[0078] E min,i ≤E i (t)≤E max,i ;

[0079] P t ≥0;

[0080] Q t ≥0;

[0081] Where: G i (t) represents the output of power generation unit i in time period t; L t Represents the load demand forecast result; G min,i represents the minimum output of power generation unit i; G max,i Indicates the maximum output of the power generation unit; E min,i represents the minimum state limit of energy storage system i; E max,i Indicates the maximum state limit of the energy storage system; E i (t) represents the state of energy storage system i in time period t.

[0082] The step 3 comprises the following steps:

[0083] Initialization: Based on the forecast results of load, power generation, equipment status, energy storage status and energy price, initialize the status of each power generation unit and energy storage system, and set the processing of each power generation unit and the charging and discharging status of the energy storage system to the minimum value that can be operated at present;

[0084] Objective function initialization: Initialize the objective function F to evaluate the pros and cons of the scheduling strategy;

[0085] Load distribution: According to the load demand prediction result L t And the output limit of the power generation unit, allocate the power generation task:

[0086] For each power generation unit i, calculate its maximum output G max,i , distribute the load according to cost efficiency, give priority to the lowest cost power generation unit, and update G in the objective function F t ;

[0087] Energy storage dispatch: E is predicted based on energy storage status t and energy price forecasts t Energy storage scheduling:

[0088] If the energy storage cost is lower than the market electricity purchase price, energy storage discharge is used first; if the market electricity price is lower than the energy storage discharge cost, energy storage charging is performed; update S in the objective function F t and B t ;

[0089] Balancing adjustment: Using constraints to adjust the processing of power generation units and energy storage systems to ensure that the constraints are met;

[0090] Objective function optimization: By adjusting G t , S t and Bt To minimize the objective function F, repeat the steps from load distribution to balance adjustment until the objective function F reaches the minimum value or satisfies the predetermined convergence condition;

[0091] Generate preliminary scheduling strategy: According to the optimized G t , S t and B t , generate a preliminary scheduling strategy.

[0092] Step 4: Based on the generated preliminary scheduling strategy, a greedy algorithm is used to optimize the preliminary scheduling strategy to obtain an optimized scheduling strategy;

[0093] The step 4 comprises the following steps:

[0094] Initialization and data input: According to the preliminary scheduling strategy, the output of the power generation unit, the energy storage status, the energy storage charge and discharge amount, and the objective function are obtained;

[0095] Load distribution of power generation units: Based on the current load demand and the unit cost of the power generation unit, the load is redistributed in accordance with the principle of minimizing the power generation cost:

[0096]

[0097] Where: represents unit fuel cost; G max,i represents the maximum output of power generation unit i; Indicates unit operation and maintenance cost;

[0098] The output of each power generation unit i According to the load demand prediction result L t and unit cost C i Adjustments are made to select the lowest cost generating unit for load distribution:

[0099]

[0100] Always: represents the power generation unit with the minimum cost in the current time period;

[0101] Energy storage system scheduling: The energy storage system scheduling strategy is based on the energy price P t and the energy storage system state S t The prediction results are used to optimize:

[0102] If the energy storage is discharged, that is, B t >0, the discharge amount is as follows:

[0103]

[0104] Where: E tIndicates the current state of the energy storage system, that is, the current stored electrical energy; Represents the discharge efficiency of the energy storage system; C max Indicates the maximum discharge power of the energy storage system;

[0105] If the energy storage is charged, that is, B t <0, the charge capacity is as follows:

[0106]

[0107] Where: Represents the charging efficiency of the energy storage system; C max Indicates the maximum charging power of the energy storage system; E max Indicates the maximum capacity of the energy storage system; S t Indicates the current energy storage status of the energy storage system, that is, the stored electrical energy;

[0108] Balance adjustment: According to load demand and equipment status, adjust the scheduling of power generation and energy storage systems to minimize the objective function F:

[0109]

[0110] Adjust by constraints and B t Guarantee the following conditions:

[0111] The total amount of power generation matches the load demand:

[0112]

[0113] The energy storage system charges and discharges to meet power limits:

[0114] -C max ≤B t ≤C max ;

[0115] Where: C max Indicates the maximum charging or discharging power limit of the energy storage system;

[0116] The energy storage state must not exceed the capacity limit:

[0117] 0≤S t ≤E max ;

[0118] Where: E max Indicates the maximum capacity of the energy storage system;

[0119] Iteration termination: When the change of the objective function F is less than the preset threshold, the algorithm is terminated and the optimized scheduling strategy is output.

[0120] In this embodiment, the power generation cost and energy storage cost can be minimized by optimizing the scheduling strategy based on the actual load demand, the unit cost of the power generation unit, and the state of the energy storage system. By redistributing the load and adjusting the energy storage charge and discharge, it can ensure that the load demand of the power grid is met at the lowest cost, thereby improving the cost-effectiveness of the scheduling strategy. This sophisticated optimization mechanism enables the system to flexibly adjust the ratio of power generation and energy storage in different time periods, further improving the accuracy of the scheduling strategy. In the load distribution process, the load is dynamically allocated according to the unit cost of the power generation unit (including fuel cost, operation and maintenance cost, etc.), and the power generation unit with the lowest cost is given priority to bear the load, thereby effectively reducing the total power generation cost. Through this flexible load distribution strategy, over-reliance on high-cost power generation units is avoided, the use of power generation resources is optimized, and the overall operation cost of the power grid is reduced.

[0121] In this embodiment, the multi-card multi-model scheduling method based on the greedy method can minimize operating costs, improve the accuracy and efficiency of scheduling strategies, and ensure the stability and safety of power grid operation by optimizing the coordination of power generation and energy storage systems, load distribution, and energy storage scheduling. By introducing dynamic optimization, flexible adjustment of parameters, and reasonable setting of constraints, the overall performance of the system is improved, making it more adaptable and scalable.

[0122] Step 5: Compare the actual operation data with the model prediction results to evaluate the performance of the scheduling strategy. According to the evaluation results, adjust the parameters of the greedy algorithm to improve the accuracy and efficiency of the scheduling strategy.

[0123] The step 5 comprises the following steps:

[0124] Define prediction error: For each important variable (such as load, power generation, energy storage status, etc.), calculate its prediction error. Assume that the actual value and the predicted value are X actual and X pred , then we can define the error of each variable ∈ t for:

[0125] ∈ t =|X actual -X pred |;

[0126] For load L t , Power generation Energy storage state S t , and energy price P t , their errors can be calculated separately:

[0127] Load forecast error:

[0128]

[0129] Where: L actual,t Indicates the actual value of the load; L pred,t Represents the load prediction value; ∈ L represents the load prediction error;

[0130] The same principle as the load forecast error calculation formula is used to calculate the power generation forecast error, energy storage state forecast error and energy price forecast error;

[0131] Scheduling strategy evaluation indicators:

[0132] Total cost: The total cost generated by the dispatch strategy, including power generation cost, energy storage charging and discharging cost, and grid power purchase cost. The goal is to minimize this cost:

[0133]

[0134] Where: F actual represents the total cost of the dispatch strategy; N represents the number of power generation units; C i represents the unit power generation cost of the i-th power generation unit; represents the actual power generation of the ith power generation unit at time t; P t represents the power purchase price of the power grid at time t; B actual,t Indicates the actual charge and discharge amount of the energy storage system at time t;

[0135] Error metric: Measure the accuracy of the scheduling policy based on the error value. For example, the mean or variance of the error can be calculated to measure the difference between the predicted results and the actual results:

[0136]

[0137] Where: MAE represents the absolute error, which is used to measure the error between the predicted value and the actual value of the scheduling strategy; T represents the length of the scheduling cycle (i.e. the total number of moments); X actual,t Indicates the actual measured value at time t (such as load, power generation, energy storage status, etc.); X pred,t Represents the predicted value at time t (given by the model).

[0138] Dispatch efficiency: The efficiency of the dispatch strategy reflects how to minimize costs while meeting power demand and equipment operation constraints under given constraints; efficiency can be quantified by the relationship between total cost and error, with lower cost and error values ​​meaning higher efficiency.

[0139] Compute optimized performance:

[0140] After comparing the generated scheduling strategy with the actual operation data, the performance of the scheduling strategy can be calculated. For example, the error (such as MAE) and total cost can be calculated to evaluate the pros and cons of the strategy.

[0141] Adjust the parameters of the greedy algorithm:

[0142] Adjust parameters based on evaluation results: Adjust the parameters in the greedy algorithm based on evaluation indicators (such as total cost, error, efficiency, etc.). There are several key parameters in the greedy algorithm that need to be adjusted:

[0143] Weight of load distribution of power generation units: According to the load demand and the output limit of the power generation unit, the dispatching strategy can be optimized by adjusting the load distribution method of the power generation unit. For example, the weight of the power generation unit in the cost function is optimized so that the power generation unit with the lowest cost gets the load first.

[0144] The adjusted load distribution rule can be implemented by modifying the weight parameters in the objective function. For example, assuming C i Represents the unit cost of the power generation unit, and the load distribution is balanced by adjusting the cost coefficient:

[0145]

[0146] Where: i represents the adjustment factor, λ i >1, increase the load distribution ratio of the power generation unit; i <1, reduce load distribution; represents the adjusted unit cost of the ith power generation unit; C i represents the original unit power generation cost of the i-th power generation unit;

[0147] Adjustment of energy storage system charging and discharging strategy: Adjust the charging and discharging priority of the energy storage system. By adjusting the priority of energy storage discharge, the energy storage system can be discharged to reduce costs when the grid electricity price is high, and charged when the grid electricity price is low.

[0148] For example, to adjust the charging / discharging priority coefficient of the energy storage system:

[0149]

[0150] Where: α charge Represents the energy storage system charging priority coefficient, which is used to adjust the charging / discharging priority; P t represents the grid electricity price at time t; P max Represents the maximum charging / discharging power of the energy storage system; ∈ P Indicates the charging / discharging efficiency of the energy storage system;

[0151] Constraints for balancing adjustment: According to the evaluation results, adjust the constraints in the balancing adjustment process. For example, if the evaluation finds that some power generation units are often in an inefficient operating state, it may be necessary to adjust the constraints to reduce the load distribution of the unit.

[0152] Set up feedback mechanism:

[0153] According to the evaluation results of the scheduling strategy, the parameters of the greedy algorithm are adjusted through the feedback mechanism. The feedback mechanism includes but is not limited to:

[0154] Adjust load distribution weights: Increase the load proportion of low-cost units by reducing the load distribution of high-cost units.

[0155] Optimize energy storage system scheduling: adjust energy storage discharge / charging strategies based on actual electricity prices and the status of the energy storage system.

[0156] Repeat the optimization process:

[0157] According to the adjusted parameters, the greedy algorithm is re-run to generate a new scheduling strategy, which is then evaluated again. Through repeated adjustments and optimizations, the scheduling strategy meets the predetermined accuracy and efficiency requirements.

[0158] Convergence conditions and termination:

[0159] When the total cost and error change of the scheduling strategy are less than the preset threshold, the optimization process is stopped and the final scheduling strategy is output. The convergence condition can be set as:

[0160] |F new -F old |<∈;

[0161] Where: ∈ represents the preset tolerance error threshold. Usually a smaller value (such as 10 -3 ), indicating that when the cost change between two iterations is small enough, the algorithm can be considered to have converged; F new represents the total cost of the scheduling strategy of the current iteration; F old Represents the total cost of the scheduling strategy for the previous iteration.

[0162] By comparing the actual operation data with the model prediction results, the accuracy and efficiency of the scheduling strategy are evaluated, and the parameters of the greedy algorithm are adjusted according to the evaluation results. The specific operations include calculating the prediction error, using the evaluation index to evaluate the scheduling strategy, adjusting the key parameters in the greedy algorithm, and continuously optimizing the scheduling strategy according to the feedback mechanism until the predetermined performance target is achieved.

[0163] The multi-card multi-model scheduling method based on the greedy method of the present invention effectively solves the technical problem of how to generate an optimized scheduling strategy and dynamically adjust the optimization parameters through a real-time feedback mechanism under the premise that there are errors and uncertainties in the prediction results of multiple models. First, by collecting and preprocessing the real-time monitoring data, historical data, meteorological data, etc. of the power grid, accurate input is provided for multi-model prediction, thereby reducing the source of prediction errors. Secondly, multiple prediction models are used to predict key variables such as load, power generation, equipment status, energy storage status and energy price to improve the accuracy of the prediction results, comprehensively consider the advantages and disadvantages of different models, and reduce the deviation that may be caused by a single model. Next, based on these prediction results, a preliminary scheduling strategy is generated using a heuristic algorithm to ensure the rationality and feasibility of the scheduling decision. On this basis, the greedy algorithm is used to optimize the preliminary scheduling strategy, and the global optimal solution is continuously approached through the local optimal selection of the greedy algorithm, so that a better scheduling effect can be achieved in the presence of prediction errors. Finally, the scheduling strategy is evaluated by comparing the difference between the actual operation data and the prediction results in real time, and the parameters of the greedy algorithm are dynamically adjusted according to the evaluation results to correct the shortcomings of the preliminary strategy. This closed-loop optimization process ensures that the dispatch strategy can adaptively respond to prediction errors and system uncertainties, improves the accuracy and efficiency of dispatch, and effectively solves the core technical problem of how to optimize power grid dispatch strategies under uncertainty.

[0164] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A multi-card multi-model scheduling method based on the greedy method, characterized in that: The following steps are involved: Step 1: Collect real-time monitoring data of the power grid, historical operation data, meteorological data, equipment parameters and energy prices, and pre-process the collected data; Step 2: Based on the preprocessed data, multiple models are used to perform predictions to obtain the prediction results of load, power generation, equipment status, energy storage status and energy price respectively; Step 3: Based on the prediction results, a heuristic algorithm is used to generate a preliminary scheduling strategy; Step 4: Based on the generated preliminary scheduling strategy, a greedy algorithm is used to optimize the preliminary scheduling strategy to obtain an optimized scheduling strategy; Step 5: Compare the actual operation data with the model prediction results to evaluate the performance of the scheduling strategy. According to the evaluation results, adjust the parameters of the greedy algorithm to improve the accuracy and efficiency of the scheduling strategy.

2. The multi-card multi-model scheduling method based on the greedy algorithm according to claim 1 is characterized in that: The objective function and constraints of the heuristic algorithm in step 3 are as follows: Objective function: Where: C t represents the power generation in the tth time period; G t represents the power generation cost in the tth time period; D t represents the energy storage discharge cost in the tth time period; S t represents the energy storage discharge in the tth time period; P t represents energy price; B t represents the amount of electricity purchased in the tth time period; Q t represents the electricity price in the tth time period; T represents the dispatching period; Constraints: G min,i ≤G i (t)≤G max,i ; E min,i ≤E i (t)≤E max,i ; P t ≥0； Q t ≥0； Where: G i (t) represents the output of power generation unit i in time period t; L t Represents the load demand forecast result; G min,i represents the minimum output of power generation unit i; G max,i Indicates the maximum output of the power generation unit; E min,i represents the minimum state limit of energy storage system i; E max,i Indicates the maximum state limit of the energy storage system; E i (t) represents the state of energy storage system i in time period t.

3. The multi-card multi-model scheduling method based on the greedy algorithm according to claim 2 is characterized in that: The step 3 comprises the following steps: Initialization: Based on the forecast results of load, power generation, equipment status, energy storage status and energy price, initialize the status of each power generation unit and energy storage system, and set the processing of each power generation unit and the charging and discharging status of the energy storage system to the minimum value that can be operated at present; Objective function initialization: Initialize the objective function F to evaluate the pros and cons of the scheduling strategy; Load distribution: According to the load demand prediction result L t And the output limit of the power generation unit, allocate the power generation task: For each power generation unit i, calculate its maximum output G max,i , distribute the load according to cost efficiency, give priority to the lowest cost power generation unit, and update G in the objective function F t ; Energy storage dispatch: E is predicted based on energy storage status t and energy price forecasts t Energy storage scheduling: If the energy storage cost is lower than the market electricity purchase price, energy storage discharge is used first; if the market electricity price is lower than the energy storage discharge cost, energy storage charging is performed; update S in the objective function F t and B t ; Balancing adjustment: Using constraints to adjust the processing of power generation units and energy storage systems to ensure that the constraints are met; Objective function optimization: By adjusting G t , S t and B t To minimize the objective function F, repeat the steps from load distribution to balance adjustment until the objective function F reaches the minimum value or meets the predetermined convergence condition; Generate preliminary scheduling strategy: According to the optimized G t , S t and B t , generate a preliminary scheduling strategy.

4. The multi-card multi-model scheduling method based on the greedy method according to claim 3 is characterized in that: The step 4 comprises the following steps: Initialization and data input: According to the preliminary scheduling strategy, the output of the power generation unit, the energy storage status, the energy storage charge and discharge amount, and the objective function are obtained; Load distribution of power generation units: Based on the current load demand and the unit cost of the power generation unit, the load is redistributed in accordance with the principle of minimizing the power generation cost: Where: represents unit fuel cost; G max,i represents the maximum output of power generation unit i; Indicates the unit operation and maintenance cost; The output of each power generation unit i According to the load demand prediction result L t and unit cost C i Adjustments are made to select the lowest cost generating unit for load distribution: Always: represents the power generation unit with the minimum cost in the current time period; Energy storage system scheduling: The energy storage system scheduling strategy is based on the energy price P t and the energy storage system state S t The prediction results are used to optimize: If the energy storage is discharged, that is, B t >0, the discharge amount is as follows: Where: E t Indicates the current state of the energy storage system, that is, the current stored electrical energy; Represents the discharge efficiency of the energy storage system; C max Indicates the maximum discharge power of the energy storage system; If energy storage is charged, that is, B t <0, the charge capacity is as follows: Where: Represents the charging efficiency of the energy storage system; C max Indicates the maximum charging power of the energy storage system; E max Indicates the maximum capacity of the energy storage system; S t Indicates the current energy storage status of the energy storage system, that is, the stored electrical energy; Balance adjustment: According to load demand and equipment status, adjust the scheduling of power generation and energy storage systems to minimize the objective function F: Adjust by constraints and B t Guarantee the following conditions: The total amount of power generation matches the load demand: The energy storage system charges and discharges to meet power limits: -C max ≤B t ≤C max ; Where: C max Indicates the maximum charging or discharging power limit of the energy storage system; The energy storage state must not exceed the capacity limit: 0≤S t ≤E max ; Where: E max Indicates the maximum capacity of the energy storage system; Iteration termination: When the change of the objective function F is less than the preset threshold, the algorithm is terminated and the optimized scheduling strategy is output.

Images