A power system day-ahead operation mode scheduling model updating method, a scheduling method and related equipment

Abstract

The application discloses a power system day-ahead operation mode scheduling model updating method and a scheduling method and related equipment, relates to the technical field of power system operation optimization, and comprises the following steps: determining a current scheduling model, an initial loss term and a physical constraint regularization term; decomposing a parameter matrix of the current scheduling model into a physical basic parameter matrix and a pseudo-correlation interference parameter matrix; taking the initial loss term and the physical constraint regularization term as joint optimization terms; constructing an optimization target based on a kernel norm of the physical basic parameter matrix and an L1 norm of the pseudo-correlation interference parameter matrix; solving the optimization target based on a parameter decoupling regularization mechanism to obtain an optimized parameter matrix; and updating the current scheduling model based on the optimized parameter matrix to obtain an updated scheduling model. The application can improve the reliability of scheduling decisions.

Description

Technical Field

[0001] This application relates to the field of power system operation optimization technology, and in particular to a method for updating the scheduling model of a power system day-ahead operation mode, a scheduling method, and related equipment. Background Technology

[0002] Machine learning-based optimization methods for day-ahead operation scheduling in power systems build optimization models by analyzing historical operating data, automatically generating decision-making schemes such as generator start-up and shutdown plans and power output allocation. However, power systems operate under normal conditions for the vast majority of the time, and their operating data, such as line power and node voltage, exhibit numerous statistically correlated but not causal pseudo-correlation characteristics. For example, in historical data, the power of a minor branch always shows a strong correlation with the power of critical sections of the system, simply because the system has been operating under similar topology and load patterns for a long time.

[0003] When the power system encounters extreme and unprecedented operating conditions, such as the first full-power commissioning of a large-scale renewable energy base accompanied by the planned shutdown of multiple main transmission lines, this spurious correlation characteristic will be completely disrupted. Because the machine learning model has never encountered this pattern during training, its internal feature weights based on spurious correlation will guide the model to make completely incorrect decisions, leading to operational risks in the power system and demonstrating the insufficient reliability of the machine learning model's decisions. Summary of the Invention

[0004] The purpose of this application is to provide a method for updating the scheduling model of day-ahead operation mode of a power system, a scheduling method, and related equipment, which can improve the reliability of scheduling decisions.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for updating the scheduling model of day-ahead operation modes of a power system, including: The current orchestration model, initial loss term, and physical constraint regularization term are determined. The current orchestration model is used to orchestrate the current day-ahead operation mode decision data based on the current operating data of the power system. The initial loss term is determined based on the historical day-ahead operation mode decision data output by the current orchestration model and the measured day-ahead operation mode decision data for the corresponding time period. The physical constraint regularization term is determined based on the basic parameters of the power system and the current operating data.

[0006] The parameter matrix of the current orchestration model is decomposed into a physical fundamental parameter matrix and a pseudo-correlation interference parameter matrix.

[0007] Using the initial loss term and the physical constraint regularization term as joint optimization terms, optimization objectives are constructed based on the kernel norm of the physical fundamental parameter matrix and the L1 norm of the pseudo-correlation interference parameter matrix, respectively. The optimization objectives are then solved based on the parameter decoupling regularization mechanism to obtain the optimized parameter matrix.

[0008] The current orchestration model is updated based on the optimized parameter matrix, and the updated current orchestration model is determined as the orchestration model.

[0009] Optionally, the optimization objective includes a first optimization objective and a second optimization objective. The first optimization objective includes the kernel norm of the physical fundamental parameter matrix, and the second optimization objective includes the L1 norm of the pseudo-correlation interference parameter matrix. The optimized parameter matrix includes the optimized pseudo-correlation interference parameter matrix and the optimized physical fundamental parameter matrix. Solving the optimization objective based on the parameter decoupling regularization mechanism yields the optimized parameter matrix, specifically including: Freeze the pseudo-correlation interference parameter matrix for the current iteration number and solve for the first optimization objective to obtain the updated physical fundamental parameter matrix for the current iteration number.

[0010] If the change between the nuclear norm of the physical fundamental parameter matrix updated in the current iteration and the nuclear norm of the physical fundamental parameter matrix updated in the previous iteration is less than a preset value, or if the current iteration reaches the preset iteration number, then the physical fundamental parameter matrix updated in the current iteration is determined as the optimized physical fundamental parameter matrix, and the optimized pseudo-correlation interference parameter matrix is ​​determined based on the optimized physical fundamental parameter matrix and the parameter matrix.

[0011] If the change between the nuclear norm of the physical fundamental parameter matrix updated in the current iteration and the nuclear norm of the physical fundamental parameter matrix updated in the previous iteration is greater than or equal to a preset value, and the current iteration has not reached the preset iteration number, then the physical fundamental parameter matrix updated in the current iteration is frozen, and the second optimization objective is solved to obtain the pseudo-correlation interference parameter matrix updated in the current iteration. Subsequently, the physical fundamental parameter matrix updated in the current iteration is determined as the physical fundamental parameter matrix for the next iteration, and the pseudo-correlation interference parameter matrix updated in the current iteration is determined as the pseudo-correlation interference parameter matrix for the next iteration. The current iteration number is incremented by 1, and the process of freezing the pseudo-correlation interference parameter matrix for the current iteration and solving the first optimization objective is returned to obtain the physical fundamental parameter matrix updated in the current iteration.

[0012] Optionally, the expression for the first optimization objective is: L+R+λ1·||W_p||*; The expression for the second optimization objective is: L+R+λ2·||W_n||1; In the formula, L is the initial loss term, R is the physical constraint regularization term, λ1 is the low-rank regularization parameter, W_p is the physical fundamental parameter matrix, ||W_p||* is the kernel norm of the physical fundamental parameter matrix, λ2 is the sparse regularization parameter, W_n is the pseudo-correlation interference parameter matrix, and ||W_n||1 is the L1 norm of the pseudo-correlation interference parameter matrix.

[0013] Optionally, the method for determining the physical constraint regularization term includes: Acquire the basic parameters and current operating data of the power system; the basic parameters include: generator set set, line set, node set, power transmission limit of each line, voltage safety upper limit and voltage safety lower limit of each node, and generator set corresponding to each node; the current operating data includes the output of each generator set, the predicted load power of each node, the predicted power transmission value of each line, and the predicted voltage amplitude of each node.

[0014] Based on the basic parameters and current operating data of the power system, calculate the power balance constraint term for each node, the line transmission capacity constraint term for each line, and the node voltage safety constraint term for each node.

[0015] The power balance constraint term of each node, the line transmission capacity constraint term of each line, and the node voltage safety constraint term of each node are normalized and summed respectively to obtain the physical constraint regularization term.

[0016] Wherein, the power balance constraint term of a node is the second norm of the algebraic sum of the injected power of the node; the injected power is generated based on the node set, the generator set set, the generator set corresponding to each node, the output of each generator set, and the load prediction power of each node.

[0017] The line transmission capacity constraint term of a line is the square of the over-limit power of the line; the over-limit power is the difference between the absolute value of the predicted transmission power and the power transmission limit of the corresponding line.

[0018] The node voltage safety constraint term for a node is the square of the deviation voltage of the node; the deviation voltage is the difference between the predicted voltage amplitude exceeding the corresponding upper limit of voltage safety, or the difference between the predicted voltage amplitude and the lower limit of voltage safety.

[0019] Optionally, the process of determining the initial loss term includes: The historical day-ahead operation mode decision data and the corresponding measured day-ahead operation mode decision data output by the current orchestration model are obtained. The measured day-ahead operation mode decision data includes the measured start-stop status of each generator set, the measured active power output setpoint of each generator set, and the measured reactive power output setpoint of each reactive power compensation device. The historical day-ahead operation mode decision data includes the predicted start-stop status of each generator set, the predicted active power output setpoint of each generator set, and the predicted reactive power output setpoint of each reactive power compensation device.

[0020] For each generator set, the mean square error between the measured active power output setpoint and the predicted active power output setpoint of the corresponding generator set is calculated and used as the first loss term for the corresponding generator set.

[0021] For each reactive power compensation device, the mean square error between the measured reactive power output setpoint and the predicted reactive power output setpoint of the corresponding reactive power compensation device is calculated and used as the second loss term of the corresponding reactive power compensation device.

[0022] The first loss term and the second loss term are normalized and weighted summed to obtain the initial loss term.

[0023] Optionally, the optimized parameter matrix includes an optimized pseudo-correlation interference parameter matrix and an optimized physical fundamental parameter matrix; updating the current orchestration model based on the optimized parameter matrix specifically includes: The optimized pseudo-correlation interference parameter matrix is ​​removed from the parameter matrix of the current orchestration model to update the current orchestration model; or, the parameter matrix of the current orchestration model is updated with the optimized physical fundamental parameter matrix to update the current orchestration model.

[0024] Secondly, this application provides a method for scheduling day-ahead operation modes of a power system, including: Obtain real-time operating data of the power system.

[0025] The real-time operating data is input into the updated orchestration model to obtain the day-ahead operating mode decision data of the power system.

[0026] The orchestration model is determined based on the orchestration model update method for the day-ahead operation mode of the power system described in any of the above-mentioned embodiments.

[0027] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the method for updating the scheduling model of the day-ahead operation mode of the power system as described above, or the steps of the method for scheduling the day-ahead operation mode of the power system.

[0028] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for updating the scheduling model of the day-ahead operating mode of a power system as described above, or the steps of the method for scheduling the day-ahead operating mode of a power system.

[0029] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method for updating the scheduling model of the day-ahead operating mode of a power system as described above, or the steps of the method for scheduling the day-ahead operating mode of a power system.

[0030] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method for updating a scheduling model of day-ahead operation modes in a power system, a scheduling method, and related equipment. In the scheduling model update method, on the one hand, the optimization objective uses the initial loss term and the physical constraint regularization term as joint optimization terms, ensuring that the optimized model still conforms to the corresponding physical constraints and maintains a low model loss, thus ensuring model stability. On the other hand, the optimization objective, constructed based on the kernel norm of the physical fundamental parameter matrix, utilizes the kernel norm to construct a low-rank regularization term during the solution process based on the parameter decoupling regularization mechanism. This determines the physical fundamental parameter matrix as a matrix solved using kernel norm regularization, approximately low-rank, with strong linear correlation between rows and columns, representing the global structured entity. Meanwhile, the L... (The sentence is incomplete and ends abruptly.) The L1 norm is used to construct the optimization objective. In the solution process based on the parameter decoupling regularization mechanism, the L1 norm is used to construct sparse regularization terms, which have element sparsity, with most elements being 0 and only a few locally non-zero terms, representing local anomalies and sparse noise. The physical fundamental parameter matrix and the pseudo-correlation interference parameter matrix are obtained by decomposing the parameter matrix of the current orchestration model. The two are subject to linear constraints in the optimization process. By solving the above optimization objective through the parameter decoupling regularization mechanism, the low-rank components and sparse components of the parameter matrix can be automatically decomposed, separating the globally stationary signal and the local discrete disturbance. This enables the identification of pseudo-correlation interference in the current orchestration model, which helps to reduce the dependence of the existing model on pseudo-correlation features in historical data and improve the reliability of orchestration decisions. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1This is an application environment diagram of a scheduling model update method for day-ahead operation modes of a power system according to an embodiment of this application; Figure 2 A flowchart illustrating a method for updating a scheduling model of a power system's day-ahead operating mode, provided as an embodiment of this application; Figure 3 A flowchart illustrating a method for scheduling day-ahead operation modes of a power system, provided as an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application.

[0033] Attached image labels: 102 terminal, 104 server. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] The power system day-ahead operation mode scheduling model update method provided in this application embodiment can be applied to, for example, Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send the current orchestration model, initial loss term, and physical constraint regularization term to be processed to server 104. After receiving the current orchestration model, initial loss term, and physical constraint regularization term, server 104 decomposes the parameter matrix of the current orchestration model into a physical fundamental parameter matrix and a pseudo-correlation interference parameter matrix. Using the initial loss term and physical constraint regularization term as joint optimization terms, it constructs optimization objectives based on the kernel norm of the physical fundamental parameter matrix and the L1 norm of the pseudo-correlation interference parameter matrix, respectively. It then solves the optimization objectives based on a parameter decoupling regularization mechanism to obtain the optimized parameter matrix. Based on the optimized parameter matrix, it updates the current orchestration model and determines the updated current orchestration model as the orchestration model. Server 104 can feed back the obtained orchestration model to terminal 102. Furthermore, in some embodiments, the orchestration model determination process can also be implemented independently by server 104 or terminal 102. For example, terminal 102 can directly perform orchestration model determination processing on the current orchestration model, initial loss term, and physical constraint regularization term to be processed; alternatively, server 104 can obtain the current orchestration model, initial loss term, and physical constraint regularization term to be processed from the data storage system and perform orchestration model determination processing on the current orchestration model, initial loss term, and physical constraint regularization term to be processed.

[0037] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0038] In one exemplary embodiment, such as Figure 2 As shown, a method for updating the scheduling model of the day-ahead operation mode of a power system is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 204. Wherein: Step 201: Determine the current orchestration model, initial loss term, and physical constraint regularization term; the current orchestration model is used to orchestrate day-ahead operation mode decision data based on the current operating data of the power system; the initial loss term is determined based on the historical day-ahead operation mode decision data output by the current orchestration model and the measured day-ahead operation mode decision data for the corresponding time period; the physical constraint regularization term is determined based on the basic parameters of the power system and the current operating data.

[0039] Step 202: Decompose the parameter matrix of the current orchestration model into a physical fundamental parameter matrix and a pseudo-correlation interference parameter matrix.

[0040] Step 203: Using the initial loss term and the physical constraint regularization term as joint optimization terms, construct optimization objectives based on the kernel norm of the physical fundamental parameter matrix and the L1 norm of the pseudo-correlation interference parameter matrix, respectively, and solve the optimization objectives based on the parameter decoupling regularization mechanism to obtain the optimized parameter matrix.

[0041] Step 204: Update the current orchestration model based on the optimized parameter matrix, and determine the updated current orchestration model as the orchestration model.

[0042] Implementing steps 201 to 204 above ensures that the optimized model still conforms to the corresponding physical constraints and maintains a low model loss, ensuring the stability of the model, realizing the automatic decomposition of the low-rank and sparse components of the parameter matrix, separating the globally stationary signal and the local discrete disturbance, realizing the identification of pseudo-correlation interference in the current orchestration model, which is conducive to solving the dependence of the existing model on pseudo-correlation features in historical data and improving the reliability of orchestration decisions.

[0043] As an example, the current orchestration model in this application is a machine learning model that can be applied to the power system industry.

[0044] This application embodiment constructs a physical constraint regularization term by combining the basic parameters of the power system with the current operating data (operating status parameters); then, by combining the historical day-ahead operating mode decision data output by the current orchestration model with the measured day-ahead operating mode decision data of the corresponding time period, a prediction loss term (i.e., initial loss term) of the current orchestration model is generated. The initial loss term and the physical constraint regularization term are combined into a joint optimization term (joint optimization objective). A parameter decoupling regularization mechanism is used to decompose the parameter matrix of the current orchestration model into a physical basic parameter matrix and a pseudo-correlation interference parameter matrix, and the two matrices are iteratively optimized to achieve the identification of pseudo-correlation interference, thereby alleviating the technical problem of insufficient decision reliability of existing machine learning models due to their reliance on pseudo-correlation features in historical data.

[0045] As an example, the method for determining the current orchestration model includes: Historical operating data of the power system and corresponding day-ahead operating mode decision data are acquired. A machine learning model is trained based on the historical operating data and the corresponding day-ahead operating mode decision data, and the trained machine learning model is determined as the current orchestration model. The machine learning model can be a deep neural network model.

[0046] The current orchestration model takes the power system's operating data as input and the corresponding day-ahead operating mode decision data as output.

[0047] Historical operating data includes unit output data, line power data, node load data, node voltage data, and topology data.

[0048] In another exemplary embodiment of this application, the method for determining the physical constraint regularization term includes: Step 11: Obtain the basic parameters and current operating data of the power system; the basic parameters include: generator set set, line set, node set, power transmission limit of each line, voltage safety upper limit and voltage safety lower limit of each node, and generator set corresponding to each node; the current operating data (operating status parameters) include the output of each generator set, the predicted load power of each node, the predicted power transmission value of each line, and the predicted voltage amplitude of each node.

[0049] As an example, the generator set set is the entire network generator set set or the target power system generator set set, including multiple generator sets; the line set is the entire network line set or the target power system line set; and the node set is the entire network node set or the target power system node set.

[0050] Step 12: Based on the basic parameters and current operating data of the power system, calculate the power balance constraint term for each node, the line transmission capacity constraint term for each line, and the node voltage safety constraint term for each node.

[0051] Step 13: Normalize and sum the power balance constraint term of each node, the line transmission capacity constraint term of each line, and the node voltage safety constraint term of each node to obtain the physical constraint regularization term.

[0052] Wherein, the power balance constraint term of a node is the second norm of the algebraic sum of the injected power of the node; the injected power is generated based on the node set, the generator set set, the generator set corresponding to each node, the output of each generator set, and the load prediction power of each node.

[0053] The line transmission capacity constraint term of a line is the square of the over-limit power of the line; the over-limit power is the difference between the absolute value of the predicted transmission power and the power transmission limit of the corresponding line.

[0054] The node voltage safety constraint term for a node is the square of the deviation voltage of the node; the deviation voltage is the difference between the predicted voltage amplitude exceeding the corresponding upper limit of voltage safety, or the difference between the predicted voltage amplitude and the lower limit of voltage safety.

[0055] The current operational data uses the predicted load power of each node, the predicted power transmission value of each line, and the predicted voltage amplitude of each node because day-ahead scheduling can only be based on predicted data. The scenario is the scheduling of day-ahead operation modes of the power system, such as preparing the unit start-up and shutdown, output allocation, and grid operation mode plan for the next 24 hours one day in advance. At the time of planning, the actual load, actual line power, and actual node voltage for the target period have not yet occurred, and there is no corresponding "real-time / measured data." The data can only rely on the results of short-term load forecasts and power flow forecasts as input.

[0056] As an example, step 12 is to construct physical constraints, including power balance constraints, line transmission capacity constraints, and node voltage safety constraints.

[0057] Step 121: Based on the node set, the generator sets corresponding to each node, the output of each generator set, and the predicted load power of each node, generate the injected power for each node, specifically as follows: The node set includes multiple nodes. For each node, based on "the generator sets corresponding to each node", all generator sets corresponding to that node are obtained, and based on "the output of each generator set", the total output of all generator sets corresponding to that node is obtained.

[0058] Obtain the predicted load power of the node based on the "predicted load power of each node".

[0059] The injected power of a node is obtained by subtracting the predicted load power of the node from the total output of all generator sets corresponding to that node, thus generating the injected power of each node.

[0060] The power balance constraint is that the algebraic sum of the injected power at each node is equal to zero.

[0061] Step 122: The line set includes multiple lines. For each line, the predicted power transmission value of that line is obtained based on the "predicted power transmission value of each line".

[0062] The absolute value of the power transmission of the line is obtained based on the predicted power transmission value of the line.

[0063] Obtain the power transmission limit for each line based on its power transmission limit.

[0064] The line transmission capacity constraint is: the absolute value of the line power transmission is less than or equal to the line power transmission limit.

[0065] Step 123: For each node, obtain the predicted voltage amplitude of that node based on the "predicted voltage amplitude of each node".

[0066] Based on the "voltage safety upper limit and voltage safety lower limit of each node", obtain the voltage safety upper limit and voltage safety lower limit of the node.

[0067] The node voltage safety constraint is: the predicted voltage amplitude of the node is within the range of the upper and lower voltage safety limits of the node.

[0068] As an example, for power balance constraints, the quadratic norm of the algebraic sum of the injected power of each node in the power balance constraint is obtained as the first regularization term of the corresponding node.

[0069] For line transmission capacity constraints, the absolute value of power transmission of each line exceeds the power transmission limit of the corresponding line by the amount, and the excess amount of each line is squared and used as the second regularization sub-item for the corresponding line.

[0070] For node voltage safety constraints, the portion of each node's voltage amplitude that exceeds the corresponding node's voltage safety upper limit or falls below the corresponding voltage safety lower limit is obtained as the deviation of each node. The deviation of each node is then squared and used as the third regularization sub-item for the corresponding node.

[0071] Normalize the first, second, and third regularization sub-terms, and then add the normalized first, second, and third regularization sub-terms together to obtain the physical constraint regularization term.

[0072] In another exemplary embodiment of this application, the process of determining the initial loss term includes: Step 14: Obtain the historical day-ahead operation mode decision data and the corresponding measured day-ahead operation mode decision data output by the current orchestration model; the measured day-ahead operation mode decision data includes the measured start-stop status of each generator set, the measured active power output setpoint of each generator set, and the measured reactive power output setpoint of each reactive power compensation device; the historical day-ahead operation mode decision data includes the predicted start-stop status of each generator set, the predicted active power output setpoint of each generator set, and the predicted reactive power output setpoint of each reactive power compensation device.

[0073] Step 15: For each generator set, calculate the mean square error between the measured active power output setpoint and the predicted active power output setpoint of the corresponding generator set, and use it as the first loss term for the corresponding generator set.

[0074] Step 16: For each reactive power compensation device, calculate the mean square error between the measured reactive power output setpoint and the predicted reactive power output setpoint of the corresponding reactive power compensation device, and use it as the second loss term of the corresponding reactive power compensation device.

[0075] Step 17: Normalize and weight the first loss term and the second loss term to obtain the initial loss term.

[0076] As an example, the specific method of step 14 includes: Obtain operational data for a historical period and the actual pre-dated operational mode decision data for the corresponding period.

[0077] Input the operational data of a historical period into the current orchestration model to obtain the historical day-ahead operational mode decision data output by the current orchestration model.

[0078] Historical pre-operational mode decision data includes the predicted start-up and shutdown status of each generator unit, the predicted active power output setpoint of each generator unit, and the predicted reactive power output setpoint of each reactive power compensation device. The predicted start-up and shutdown status of a generator unit is represented by a vector of 0s and 1s, where 1 indicates the generator unit is predicted to be in operation during the corresponding time period, and 0 indicates the generator unit is predicted to be out of operation during the corresponding time period. The predicted active power output setpoint of each generator unit is a real-number vector representing the predicted planned power generation value of each generator unit in each time period of the next scheduling cycle, in megawatts (MW). The predicted reactive power output setpoint of each reactive power compensation device is a real-number vector representing the predicted planned reactive power output or absorption value of each device in each time period of the next scheduling cycle, in megavars (Mvars).

[0079] In another exemplary embodiment of this application, step 202 decomposes the parameter matrix of the current orchestration model into a linear combination of the physical fundamental parameter matrix to be solved and the pseudo-correlation interference parameter matrix to be solved, wherein the linear combination is matrix addition. This can be understood as follows: let the parameter matrix be W, and let W = W_p + W_n, where W_p is the physical fundamental parameter matrix to be solved, and W_n is the pseudo-correlation interference parameter matrix to be solved.

[0080] In another exemplary embodiment of this application, the optimization objective includes a first optimization objective and a second optimization objective, wherein the expression for the first optimization objective is: L+R+λ1·||W_p||*; The expression for the second optimization objective is: L+R+λ2·||W_n||1; In the formula, L is the initial loss term, R is the physical constraint regularization term, λ1 is the low-rank regularization parameter (λ1 > 0), W_p is the physical fundamental parameter matrix, ||W_p||* is the kernel norm of the physical fundamental parameter matrix, λ2 is the sparse regularization parameter (λ2 > 0), W_n is the pseudo-correlation interference parameter matrix, and ||W_n||1 is the L1 norm of the pseudo-correlation interference parameter matrix.

[0081] In this embodiment, a low-rank regularization term is constructed using the nuclear norm of the physical fundamental parameter matrix to be solved; the low-rank regularization term is used to constrain the final physical fundamental parameter matrix W_p to have low-rank characteristics.

[0082] As an explanation, a low-rank regularization term is constructed using the nuclear norm of the physical fundamental parameter matrix W_p to be solved. This means: let the nuclear norm of the physical fundamental parameter matrix W_p to be solved be ||W_p||*; let the first regularization parameter be λ1, λ1>0, then the low-rank regularization term is λ1·||W_p||*.

[0083] On the other hand, a sparse regularization term is constructed using the L1 norm of the pseudo-correlation interference parameter matrix to be solved. The sparse regularization term is used to constrain the final pseudo-correlation interference parameter matrix W_n to have sparse properties.

[0084] As an explanation, a sparse regularization term is constructed using the L1 norm of the pseudo-correlation interference parameter matrix W_n to be solved. This means: let the L1 norm of the pseudo-correlation interference parameter matrix W_n to be solved be ||W_n||1; let the second regularization parameter be λ2, λ2>0, then the sparse regularization term is λ2·||W_n||1.

[0085] In another exemplary embodiment of this application, the optimized parameter matrix includes an optimized pseudo-correlation interference parameter matrix and an optimized physical fundamental parameter matrix; the optimization objective can be iteratively solved using the alternating direction multiplier method. Therefore, step 203, which solves the optimization objective based on the parameter decoupling regularization mechanism to obtain the optimized parameter matrix, specifically includes: Step 31: Freeze the pseudo-correlation interference parameter matrix for the current iteration number and solve for the first optimization objective to obtain the updated physical fundamental parameter matrix for the current iteration number. As an example, the first optimization objective can be minimized using gradient descent, meaning we need to find a W_p such that the overall value of L+R+λ1·||W_p||* is minimized.

[0086] Step 32: If the change between the nuclear norm of the physical fundamental parameter matrix updated in the current iteration number and the nuclear norm of the physical fundamental parameter matrix updated in the previous iteration number is less than a preset value, or if the current iteration number reaches the preset iteration number, then the physical fundamental parameter matrix updated in the current iteration number is determined as the optimized physical fundamental parameter matrix, and the optimized pseudo-correlation interference parameter matrix is ​​determined based on the optimized physical fundamental parameter matrix and the parameter matrix.

[0087] If the change between the nuclear norm of the physical fundamental parameter matrix updated in the current iteration and the nuclear norm of the physical fundamental parameter matrix updated in the previous iteration is greater than or equal to a preset value, and the current iteration has not reached the preset iteration number, then the physical fundamental parameter matrix updated in the current iteration is frozen, and the second optimization objective is solved to obtain the pseudo-correlation interference parameter matrix updated in the current iteration. Then, the physical fundamental parameter matrix updated in the current iteration is determined as the physical fundamental parameter matrix for the next iteration, and the pseudo-correlation interference parameter matrix updated in the current iteration is determined as the pseudo-correlation interference parameter matrix for the next iteration. The current iteration number is incremented by 1, and the process returns to step 31.

[0088] As an example, the second optimization objective can be minimized using the proximal gradient method, which means finding a W_n such that the overall value of L+R+λ2·||W_n||1 is minimized.

[0089] In another exemplary embodiment of this application, the optimized parameter matrix includes an optimized pseudo-correlation interference parameter matrix and an optimized physical fundamental parameter matrix; updating the current orchestration model based on the optimized parameter matrix specifically includes: Step 41: Remove the optimized pseudo-correlation interference parameter matrix from the parameter matrix of the current orchestration model to update the current orchestration model; or, update the parameter matrix of the current orchestration model with the optimized physical fundamental parameter matrix to update the current orchestration model.

[0090] In another exemplary embodiment, a 2×2 parameter matrix is ​​used as an example (this is just illustrative data; it is not actual matrix data, as neural network parameters have many dimensions, and the 2×2 parameter matrix is ​​not actually calculated). ; W = W_p + W_n; Initialize W_n to : ; At this time, W_p is : .

[0091] During the first iteration, fix Construct the first optimization objective: L + R + λ1·||W_p||*, where After updating via gradient descent, we get: .

[0092] At this point, the low-rank constraint will push the matrix toward a more compact structure.

[0093] After the subsequent fixed update Construct the second optimization sub-objective: L + R + λ²·||W_n||¹, where, After proximal gradient update, we get: .

[0094] Due to the L1 norm constraint, only a small number of non-zero elements are retained.

[0095] The above steps are repeated during the second iteration.

[0096] fixed : .

[0097] renew The result was: .

[0098] Updated later get: .

[0099] After multiple iterations, the convergence condition is met. Let's assume that at this point: ; .

[0100] Then, proceed to step 41, discard W_n, and obtain the orchestration model.

[0101] Alternatively, you can directly use W_p, that is... This serves as the parameter matrix for the final model, resulting in the orchestration model.

[0102] As an example, the update mechanism of the orchestration model in the above embodiments can be: I. Periodic Automatic Update Mechanism: Updates are automatically triggered according to a preset fixed period (such as daily or weekly, as the matching day-to-day scheduling business is usually executed on a daily / weekly basis), without the need for manual initiation.

[0103] The newly accumulated historical operating data of the power system during the cycle, and the actual day-ahead operating mode decision data for the corresponding period; if the basic parameters of the power grid (units, lines, voltage limits, etc.) are adjusted during the cycle, the latest basic parameters are included simultaneously.

[0104] The execution process fully reuses the entire process of the orchestration model update method of this invention: the currently used orchestration model is used as the object of the new round of updates (the current orchestration model).

[0105] Based on the newly added historical operating data and measured decision data, the initial loss term for the new round is calculated; based on the latest power grid basic parameters and current operating data, the physical constraint regularization term is calculated; the parameter matrix decomposition (split into physical basic parameter matrix + pseudo-correlation interference parameter matrix) and the parameter decoupling optimization process of alternating freezing and iterative solution are re-executed; the optimized parameter matrix is ​​obtained, the model parameters are updated, and the periodic iteration is completed; the purpose is to continuously absorb new operating data, gradually remove outdated pseudo-correlation features, and ensure that the model always adapts to the current power grid's normal operating mode.

[0106] II. Manual Intervention Triggered Update Mechanism: This mechanism is used to respond to major changes in the power grid or special scenarios. It is a special update that is manually triggered and is considered an unconventional adjustment.

[0107] 1. Significant changes to the power grid topology / parameters: such as the addition / retirement of generating units, commissioning / shutdown of main lines, adjustment of node voltage safety limits, and revision of line power transmission limits; 2. In the event of extreme operating scenarios: such as the commissioning of ultra-large-scale new energy bases or changes in operating modes after major power grid failures, the reliability of the original model's decision-making decreases; 3. Business rule adjustments: such as updates to scheduling security standards and changes to operational constraint policies.

[0108] The execution process involves manual triggering of model retraining by operations and maintenance personnel. The core process still follows the orchestration model update method of this invention, while also supporting manual adjustment and optimization parameters. Input the updated power grid basic parameters, the target scenario's operational data, and the measured decision data; The values ​​of the low-rank regularization parameter λ1 and the sparse regularization parameter λ2 can be manually adjusted to adapt to the requirements of physical constraint strength and pseudo-correlation feature removal in different scenarios. After completing the parameter decoupling optimization solution, an orchestration model adapted to the new scenario is generated. After manual verification, the original online model is replaced.

[0109] All the model updates in the above modes strictly follow the core technical logic of this invention: always take "initial loss term + physical constraint regularization term" as the joint optimization objective, and separate physical laws and pseudo-correlation interference through low-rank-sparse decomposition of parameter matrix to ensure that the updated model satisfies the physical constraints of the power grid and continuously reduces the dependence on pseudo-correlation features.

[0110] In one exemplary embodiment, such as Figure 3 As shown, a method for scheduling day-ahead operation modes of a power system is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 901 to 902. Wherein: Step 901: Obtain real-time operating data of the power system.

[0111] Step 902: Input the real-time operating data into the updated orchestration model to obtain the day-ahead operating mode decision data of the power system.

[0112] The updated orchestration model is determined based on the orchestration model update method for day-ahead operation of the power system described in any of the above embodiments.

[0113] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown. The computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores process data for determining the orchestration model or for orchestrating the day-ahead operating mode of the power system. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for updating a day-ahead operating mode orchestration model of the power system, or a method for orchestrating the day-ahead operating mode of the power system.

[0114] Those skilled in the art will understand that Figure 4The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0115] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0116] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0117] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant regulations and be authorized by the owner of the corresponding device.

[0118] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0119] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0120] The embodiments provided in this application should be understood to be representative of systems or methods that can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may exist in actual implementation.

[0121] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0122] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0123] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0124] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for updating the scheduling model of a power system's day-ahead operating mode, characterized in that, The orchestration model update method includes: The current orchestration model, initial loss term, and physical constraint regularization term are determined. The current orchestration model is used to orchestrate day-ahead operation mode decision data based on the current operating data of the power system. The initial loss term is determined based on the historical day-ahead operation mode decision data output by the current orchestration model and the measured day-ahead operation mode decision data for the corresponding time period. The physical constraint regularization term is determined based on the basic parameters of the power system and the current operating data. The parameter matrix of the current orchestration model is decomposed into a physical fundamental parameter matrix and a pseudo-correlation interference parameter matrix; Using the initial loss term and the physical constraint regularization term as joint optimization terms, optimization objectives are constructed based on the kernel norm of the physical fundamental parameter matrix and the L1 norm of the pseudo-correlation interference parameter matrix, respectively. The optimization objectives are then solved based on the parameter decoupling regularization mechanism to obtain the optimized parameter matrix. The current orchestration model is updated based on the optimized parameter matrix to obtain the updated orchestration model.

2. The method for updating the scheduling model of day-ahead operation mode of a power system according to claim 1, characterized in that, The optimization objectives include a first optimization objective and a second optimization objective. The first optimization objective includes the kernel norm of the physical fundamental parameter matrix, and the second optimization objective includes the L1 norm of the pseudo-correlation interference parameter matrix. The optimized parameter matrix includes the optimized pseudo-correlation interference parameter matrix and the optimized physical fundamental parameter matrix. The optimization objective is solved based on the parameter decoupling regularization mechanism, resulting in the optimized parameter matrix, which includes: Freeze the pseudo-correlation interference parameter matrix for the current iteration number and solve for the first optimization objective to obtain the physical fundamental parameter matrix updated for the current iteration number; If the change between the nuclear norm of the physical fundamental parameter matrix updated in the current iteration and the nuclear norm of the physical fundamental parameter matrix updated in the previous iteration is less than a preset value, or if the current iteration reaches the preset iteration number, then the physical fundamental parameter matrix updated in the current iteration is determined as the optimized physical fundamental parameter matrix, and the optimized pseudo-correlation interference parameter matrix is ​​determined based on the optimized physical fundamental parameter matrix and the parameter matrix. If the change between the nuclear norm of the physical fundamental parameter matrix updated in the current iteration and the nuclear norm of the physical fundamental parameter matrix updated in the previous iteration is greater than or equal to a preset value, and the current iteration has not reached the preset iteration number, then the physical fundamental parameter matrix updated in the current iteration is frozen, and the second optimization objective is solved to obtain the pseudo-correlation interference parameter matrix updated in the current iteration. Subsequently, the physical fundamental parameter matrix updated in the current iteration is determined as the physical fundamental parameter matrix for the next iteration, and the pseudo-correlation interference parameter matrix updated in the current iteration is determined as the pseudo-correlation interference parameter matrix for the next iteration. The current iteration number is incremented by 1, and the process of freezing the pseudo-correlation interference parameter matrix for the current iteration and solving the first optimization objective is returned to obtain the physical fundamental parameter matrix updated in the current iteration.

3. The method for updating the scheduling model of day-ahead operation mode of a power system according to claim 2, characterized in that, The expression for the first optimization objective is: L+R+λ1·||W_p||*; The expression for the second optimization objective is: L+R+λ2·||W_n||1; In the formula, L is the initial loss term, R is the physical constraint regularization term, λ1 is the low-rank regularization parameter, W_p is the physical fundamental parameter matrix, ||W_p||* is the kernel norm of the physical fundamental parameter matrix, λ2 is the sparse regularization parameter, W_n is the pseudo-correlation interference parameter matrix, and ||W_n||1 is the L1 norm of the pseudo-correlation interference parameter matrix.

4. The method for updating the scheduling model of day-ahead operation mode of a power system according to claim 1, characterized in that, The method for determining the physical constraint regularization term includes: Acquire the basic parameters and current operating data of the power system; the basic parameters include: generator set set, line set, node set, power transmission limit of each line, voltage safety upper limit and voltage safety lower limit of each node, and generator set corresponding to each node; the current operating data includes the output of each generator set, the predicted load power of each node, the predicted power transmission value of each line, and the predicted voltage amplitude of each node. Based on the basic parameters and current operating data of the power system, calculate the power balance constraint term for each node, the line transmission capacity constraint term for each line, and the node voltage safety constraint term for each node. The power balance constraint term of each node, the line transmission capacity constraint term of each line, and the node voltage safety constraint term of each node are normalized and summed respectively to obtain the physical constraint regularization term. Wherein, the power balance constraint term of a node is the second norm of the algebraic sum of the injected power of the node; the injected power is generated based on the node set, the generator set set, the generator set corresponding to each node, the output of each generator set, and the load prediction power of each node. The line transmission capacity constraint term of a line is the square of the over-limit power of the line; the over-limit power is the difference between the absolute value of the predicted transmission power and the power transmission limit of the corresponding line. The node voltage safety constraint term for a node is the square of the deviation voltage of the node; the deviation voltage is the difference between the predicted voltage amplitude exceeding the corresponding upper limit of voltage safety, or the difference between the predicted voltage amplitude and the lower limit of voltage safety.

5. The method for updating the scheduling model of day-ahead operation mode of a power system according to claim 1, characterized in that, The process of determining the initial loss term includes: The system acquires the historical day-ahead operation mode decision data and the corresponding measured day-ahead operation mode decision data output by the current orchestration model. The measured day-ahead operation mode decision data includes the measured start-stop status of each generator set, the measured active power output setpoint of each generator set, and the measured reactive power output setpoint of each reactive power compensation device. The historical day-ahead operation mode decision data includes the predicted start-stop status of each generator set, the predicted active power output setpoint of each generator set, and the predicted reactive power output setpoint of each reactive power compensation device. For each generator set, the mean square error between the measured active power output setpoint and the predicted active power output setpoint of the corresponding generator set is calculated and used as the first loss term for the corresponding generator set. For each reactive power compensation device, the mean square error between the measured reactive power output setpoint and the predicted reactive power output setpoint of the corresponding reactive power compensation device is calculated and used as the second loss term of the corresponding reactive power compensation device. The first loss term and the second loss term are normalized and weighted summed to obtain the initial loss term.

6. The method for updating the scheduling model of day-ahead operation mode of a power system according to claim 1, characterized in that, The optimized parameter matrix includes the optimized pseudo-correlation interference parameter matrix and the optimized physical fundamental parameter matrix; The current orchestration model is updated based on the optimized parameter matrix, specifically including: The optimized pseudo-correlation interference parameter matrix is ​​removed from the parameter matrix of the current orchestration model to update the current orchestration model; or, the parameter matrix of the current orchestration model is updated with the optimized physical fundamental parameter matrix to update the current orchestration model.

7. A method for scheduling day-ahead operation modes of a power system, characterized in that, The day-ahead operation mode scheduling method for power systems includes: Obtain real-time operational data of the power system; The real-time operational data is input into the updated orchestration model to obtain the day-ahead operation mode decision data of the power system. The updated orchestration model is determined based on the orchestration model update method for day-ahead operation of the power system as described in any one of claims 1-6.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the power system day-ahead operation mode scheduling model update method according to any one of claims 1-6, or the power system day-ahead operation mode scheduling method according to claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for updating the day-ahead operation mode of a power system as described in any one of claims 1-6, or the method for arranging day-ahead operation modes of a power system as described in claim 7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for updating the day-ahead operation mode of a power system as described in any one of claims 1-6, or the method for arranging day-ahead operation modes of a power system as described in claim 7.

Images