Multi-module integrated power grid investment scale dynamic optimization decision-making system and method

Abstract

The invention relates to the field of power system automatic decision making, in particular to a multi-module integrated power grid investment scale dynamic optimization decision making system and method, which comprises a data acquisition unit, an investment prediction analysis module, an investment efficiency analysis module, a safety margin analysis module, an investment balance analysis module, a data storage unit and a man-machine interaction unit, the data acquisition unit is butted with multiple data sources of the power system to automatically acquire indexes; the investment prediction module outputs a total amount prediction value through a multiple linear regression model; the efficiency analysis module adopts a four-stage DEA model to realize unbiased efficiency evaluation; the safety margin module dynamically quantifies the safety state of the power grid; the investment balancing module serves as a core engine to execute multi-objective optimization distribution; the data storage unit supports full-process data backup, the man-machine interaction unit realizes visual operation and scheme export, investment decision integration, dynamics and precision are realized, and decision efficiency and comprehensive benefits are improved.

Description

Technical Field

[0001] This invention relates to the field of power system automation decision-making, specifically to a multi-module integrated dynamic optimization decision-making system and method for power grid investment scale. Background Technology

[0002] The new power system exhibits characteristics of diversified power source structure, complex load characteristics, and dynamic operating status, placing stringent demands on the scientific rigor, accuracy, and coordination of power grid investment decisions. Existing power grid investment decision-making technologies have significant shortcomings:

[0003] First, the decision-making process is highly fragmented. Investment forecasting, efficiency assessment, security analysis, and regional allocation are handled by different tools or departments. There is a lack of standardized data interfaces and collaborative computing mechanisms. Data collection relies on manual integration, and data silos are formed between modules, resulting in a disconnect between the total forecast and the micro-allocation, and a broken decision-making chain.

[0004] Secondly, the accuracy of prediction and assessment is insufficient. Investment prediction models are static and fixed, and do not adaptively update with the evolution of the power grid, resulting in large prediction errors; efficiency assessments do not effectively isolate environmental factors and random errors, making it difficult to reflect the true management level of the region; safety analysis relies on static thresholds, making it impossible to quantify real-time safety margins and the safety benefits of investment.

[0005] Furthermore, there is a lack of multi-objective coordination capabilities. The allocation process separates the objectives of efficiency, safety, and balance, relying on subjective manual weighing and lacking a unified and quantifiable multi-objective optimization framework, making it difficult to achieve comprehensive optimality.

[0006] Finally, operation and maintenance are inconvenient. The decision-making process lacks visualization, data storage is scattered and lacks a backup mechanism, and adjustments to the plan require repetitive manual operations, resulting in low efficiency. Summary of the Invention

[0007] (a) Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a multi-module integrated dynamic optimization decision-making system and method for power grid investment scale.

[0009] (II) Technical Solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: The multi-module integrated dynamic optimization decision-making system for power grid investment scale includes a data acquisition unit, an investment forecasting and analysis module, an investment efficiency analysis module, a safety margin analysis module, an investment equilibrium analysis module, a data storage unit, and a human-computer interaction unit. Each component is connected to an internal bus via a standardized data interface. The data acquisition unit interfaces with the dispatch automation system, infrastructure management system, production management system, and SCADA system via a dedicated power system communication protocol, automatically collecting power grid investment-related indicators, efficiency evaluation indicators, and safe operation indicators. The investment forecasting and analysis module is communicatively connected to the data acquisition unit. The investment efficiency analysis module is also communicatively connected to the data acquisition unit, employing a four-stage DEA model for unbiased efficiency evaluation. The safety margin analysis module is communicatively connected to the data acquisition unit. The investment equilibrium analysis module serves as the core optimization engine, communicating with the investment forecasting and analysis module, the investment efficiency analysis module, and the safety margin analysis module.

[0011] Preferably, the built-in multiple linear regression prediction model is trained using the least squares method, with the following formula:

[0012] I t =α0+Σα it x it +ε

[0013] Among them, I t As a predictor variable, x it The variables collected are α0 and α. it ε represents the influence coefficient, and ε represents the error term. The module incorporates a leave-one-out cross-validation mechanism to evaluate the model's robustness and outputs the predicted total power grid investment value I for the next t+n periods. t+n The data is then transmitted to the investment equilibrium analysis module.

[0014] More preferably, the investment efficiency analysis module includes the following four stages:

[0015] Phase 1: Preliminary calculation of the original efficiency value based on the BCC model, with the model constraint being: minθ k

[0016]

[0017] Where, θ k x represents the pure technical efficiency value of the k-th DMU; ij y represents the i-th type of input for the j-th project; mj Let r represent the m-th output of the j-th project; r and t represent the number of input and output variables of the project, respectively.

[0018] Phase Two: Constructing an SFA regression model to separate environmental factors from random error, the formula is: Srk =f(Z) k ;β r )+V rk +U rk , among which, S rk f(Z) represents the slack variable of the nth input of the k-th DMU; k ;β r ) represents the effect of external environmental factors on the slack variable S rk The influence, take f(Z) k ;β r ) = Z k ·β r Z k β represents observable environmental factor variables. r Z represents the environmental factor variable k The corresponding parameter vector that needs to be estimated; V rk +U rk V represents the mixed error term. rk For the random error term, when it follows a normal distribution, U rk To manage inefficiency terms, when they conform to a truncated normal distribution, V rk with U rk Independent and uncorrelated, parameters are estimated using the maximum likelihood estimation method, and the input data are adjusted to homogeneous environments and random states;

[0019] Then, the maximum likelihood estimation method was used to estimate β. r σ 2 And the γ parameter, then calculate the management inefficiency term U. rk :

[0020]

[0021] Finally, the formula was adjusted to estimate β. r σ 2 and γ parameter values ​​and management inefficiency terms U rk The random error term V is calculated based on this. rk By adjusting the input values ​​of different decision-making units, they can be placed under homogeneous environmental and luck conditions:

[0022]

[0023] Among them, X rk For the actual input of the decision-making unit, To adjust the input, [max(Z) k ·β r )-Z k ·β r ] indicates adjusting different decision-making units to homogeneous environmental conditions, [max(V rk )-Vrk This indicates adjusting different decision-making units to the same state of nature;

[0024] Phase 3: Import the input values ​​of the decision-making units after the adjustments in Phase 2 into the DEA model in Phase 1 to recalculate the relative R&D efficiency of each decision-making unit;

[0025] Phase 4: Substitute the adjusted data into the BCC model, recalculate the pure technical efficiency, overall efficiency, and scale efficiency, and output the unbiased efficiency values ​​for each region to the investment equilibrium analysis module.

[0026] Preferably, the safety margin analysis module includes:

[0027] Safety margin quantification: Define the power grid operation safety domain, calculate the weighted distance from the current operating state point to the boundary of the safety domain, and obtain the real-time safety margin s for each region. i ;

[0028] Investment safety margin modeling: Based on historical data regression analysis or power grid digital simulation, establish a functional relationship f between investment and safety margin improvement. i (I i ), where I i For regional investment quotas;

[0029] Output: Current safety margin s for each region i and investment safety gain function f i (I i The data is then transmitted to the investment equilibrium analysis module.

[0030] Preferably, the investment equilibrium analysis module includes:

[0031] Initial allocation: Investment is allocated based on the comprehensive efficiency value of each region, using the following formula: Among them, I i For the investment allocation of unit i in year t, I t Let t be the total investment amount in year t. The percentage of investment efficiency for unit i in the previous year;

[0032] Balance calculation: Plot the Lorenz curve and use the trapezoidal area method to calculate the Gini coefficient G of the unit investment increase in electricity sales and capacity ratio. j ;

[0033] Weight determination: The weights ω of each equilibrium index are calculated using the entropy method. j ;

[0034] Multi-objective optimization: An optimization model is constructed with the objective of minimizing the weighted sum of the Gini coefficients, and conditions including total investment constraints, regional investment adjustment ratio constraints, and safety margin lower bound constraints.

[0035] Objective function:

[0036] Constraints:

[0037] Total investment adjustment ratio constraint:

[0038] Constraints on the adjustment ratio of investment amount for each unit:

[0039] Dynamic safety constraints: s i +f i (I i )≥s min ;

[0040] Gini coefficient constraints for each indicator: G 1(j) ≤G 0(j) ;

[0041] In the formula, F is the weighted average of the Gini coefficients of each indicator, j is the number of the three indicators, i is the allocation unit number, and W... 0(i) W 1(i) These represent the investment amounts for each unit before and after the adjustment, respectively. q0 and q1 are the upper and lower limits of the total investment adjustment ratio, respectively. i0 and p i1 G sets the upper and lower limits for adjusting the investment amount of each unit. 0(j) Let j be the initial value of the Gini coefficient corresponding to index j;

[0042] Output: Final power grid investment allocation plan.

[0043] A further preferred method for dynamic optimization of power grid investment scale through multi-module integration includes the following steps:

[0044] S1. Data Acquisition: The data acquisition unit connects to various data sources in the power system to automatically collect the raw data required for investment forecasting, efficiency assessment, and safety analysis, and performs data cleaning and format standardization processing.

[0045] S2. Total Investment Forecast: The investment forecasting and analysis module calls a multiple linear regression model, inputs standardized data, and outputs the predicted total investment in the future power grid after passing the leave-one-out cross-validation. t+n ;

[0046] S3. Unbiased Efficiency Assessment: The investment efficiency analysis module uses a three-stage DEA model to sequentially complete the original efficiency calculation, separation of environmental and random errors, adjustment of input data, and efficiency reassessment, outputting the unbiased efficiency value for each region; S4. Safety Margin Diagnosis: The safety margin analysis module quantifies the current safety margin s for each region based on real-time operational data. i Establish the investment safety gain function f i (Ii );

[0047] S5. Multi-objective equilibrium optimization: The investment equilibrium analysis module receives the output data from the preceding module, first completes the initial investment allocation based on the efficiency value, and then constructs and solves the multi-objective optimization model by using the Lorenz curve, Gini coefficient calculation, and entropy method weight determination, and outputs the final investment allocation scheme.

[0048] S6. Solution Output and Feedback: The investment solution is displayed and exported through the human-computer interaction unit, and the data storage unit stores the relevant data synchronously; if adjustments are needed, return to S1-S5 to re-execute the optimization.

[0049] Preferably, the power grid investment-related indicators include historical investment amounts, load growth data, and power supply structure parameters; the efficiency evaluation indicators include input-output data and environmental variable data; and the safe operation indicators include capacity-to-load ratio, load factor, and power flow distribution data.

[0050] Preferably, the investment forecasting and analysis module supports adaptive model updates, automatically calling the latest historical data to retrain the parameters of the multiple linear regression model every quarter;

[0051] In the SFA regression model of the investment efficiency analysis module, environmental variables include regional economic level, power grid scale, and climate conditions. The random error term adopts the assumptions of normal distribution and truncated normal distribution to improve the accuracy of error separation.

[0052] The safety domain model of the safety margin analysis module supports custom configuration. The safety constraint parameters can be adjusted according to different voltage levels of 220kV or 110kV. The weights in the weighted distance calculation are determined by the analytic hierarchy process.

[0053] The optimization model of the investment equilibrium analysis module is solved using an improved particle swarm optimization algorithm, and a dynamic adjustment strategy for inertia weight is introduced.

[0054] The safety margin constraint of the investment balance analysis module supports dynamic adjustment, and automatically adjusts the value of smin according to the early warning level of the power grid operation status.

[0055] Preferably, the data storage unit adopts an industrial-grade database to store the collected raw data, intermediate calculation results of each module, optimized model parameters and final investment plan, and supports local data backup and cloud synchronization. The data acquisition unit supports IEC61850 and DL / T860 protocols, with a data sampling frequency of 15 minutes / time, and has a built-in edge computing preprocessing function to perform noise reduction and interpolation processing on the raw data.

[0056] Preferably, the human-computer interaction unit includes a touch screen and a remote client, supporting investment parameter configuration, model parameter adjustment, visualization of the decision-making process, and export of investment plans.

[0057] (III) Beneficial Effects

[0058] Compared with existing technologies, this invention provides a multi-module integrated dynamic optimization decision-making system and method for power grid investment scale, which has the following beneficial effects:

[0059] This technical solution automatically connects to scheduling, infrastructure, production, and SCADA systems through a data acquisition unit, enabling high-frequency, standardized acquisition of three types of indicators: investment, efficiency, and safety. It also supports edge preprocessing to ensure data quality.

[0060] The investment forecasting and analysis module uses a multiple linear regression model with cross-validation to dynamically output the total future investment amount; the investment efficiency analysis module innovatively introduces a four-stage DEA-SFA hybrid model to effectively isolate environmental and random interference and obtain unbiased efficiency values ​​for each region; the safety margin analysis module quantifies the operating status and the distance to the safety boundary and establishes an investment safety gain function to support risk-controlled investment allocation.

[0061] The core equilibrium optimization engine aims to minimize the weighted Gini coefficient, integrating three principles: efficiency priority, balanced development, and safety baseline. It constructs a multi-objective optimization model with total investment constraints, unit adjustment limits, and dynamic safety thresholds, employing an improved particle swarm optimization algorithm for efficient solution, ensuring the solution is economical, fair, and reliable. The human-computer interaction unit supports full-process visualization and parameter tuning, while the data storage unit achieves closed-loop data management across the entire chain.

[0062] The entire system can adaptively update forecasting and efficiency models quarterly, supports security domain configurations for different voltage levels, and is suitable for medium- and long-term investment planning of provincial or municipal power grids. Compared with traditional experience-based allocation methods, this invention can improve investment efficiency, reduce regional development imbalances, and ensure that security margins in key areas meet standards, providing intelligent and refined decision support for the high-quality development of new power systems. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the technical architecture and workflow of the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] Please see Figure 1 The present invention discloses a multi-module integrated dynamic optimization decision-making system for power grid investment scale, comprising a data acquisition unit, an investment forecasting and analysis module, an investment efficiency analysis module, a safety margin analysis module, an investment equilibrium analysis module, a data storage unit, and a human-computer interaction unit. Each component is connected to an internal bus via a standardized data interface. The data acquisition unit interfaces with the dispatch automation system, infrastructure management system, production management system, and SCADA system via a dedicated power system communication protocol, automatically collecting power grid investment-related indicators, efficiency evaluation indicators, and safe operation indicators. The investment forecasting and analysis module communicates with the data acquisition unit. The investment efficiency analysis module communicates with the data acquisition unit and employs a four-stage DEA model for unbiased efficiency evaluation. The safety margin analysis module communicates with the data acquisition unit. The investment equilibrium analysis module serves as the core optimization engine and communicates with the investment forecasting and analysis module, the investment efficiency analysis module, and the safety margin analysis module.

[0066] The working principle of this technical solution is to construct a fully automated decision-making system encompassing data acquisition, multi-module collaborative analysis, multi-objective equilibrium optimization, and solution output. It achieves unified acquisition and preprocessing of multi-source heterogeneous data through standardized data interfaces. Utilizing three modules—investment forecasting, efficiency assessment, and safety diagnosis—it outputs total investment constraints, unbiased efficiency benchmarks, and hard safety constraints, respectively. Finally, the investment equilibrium analysis module integrates multi-dimensional inputs and, through mathematical modeling and intelligent solving, generates a power grid investment solution that satisfies the requirements of controllable total investment, optimal efficiency, safety compliance, and balanced allocation, thus completely solving the problems of fragmented and static traditional decision-making.

[0067] The data acquisition unit acts as a data interaction hub, connecting with multiple data sources such as dispatch automation systems, infrastructure management systems, production management systems, and SCADA systems via power system-specific communication protocols, such as IEC61850 and DL / T860. It automatically collects three types of core indicators at a sampling frequency of 15 minutes per instance: power grid investment-related indicators, such as historical investment amounts, load growth data, and power supply structure parameters; efficiency evaluation indicators, such as input-output data and environmental variable data; and safe operation indicators, such as capacity-to-load ratio, load factor, and power flow distribution data. It incorporates built-in edge computing preprocessing functions, employing mean filtering algorithms for noise reduction and linear interpolation to complete missing data. The raw data is standardized into a unified format, such as CSV, and transmitted to various analysis modules via the internal bus, while simultaneously being backed up to the data storage unit, ensuring data real-time performance, integrity, and consistency.

[0068] The data storage unit utilizes industrial-grade databases, such as MySQL / PostgreSQL, and constructs a three-tiered storage architecture encompassing raw data, intermediate results, and the final solution. The raw data layer stores standardized data preprocessed by the acquisition unit; the intermediate results layer stores calculation process data from each module, such as regression model parameters, efficiency calculations, and safety margin data; and the final solution layer stores the optimized investment allocation plan. It supports both local hard drive backup and cloud synchronization, employs RAID redundancy technology to prevent data loss, and maintains a data read / write latency of ≤1 second, providing data support for module interaction, solution traceability, and secondary optimization.

[0069] The human-computer interaction unit consists of a touch screen and a remote client. It communicates with the system's core module via Ethernet. The touch screen provides a local operating interface, supporting investment parameter configuration, such as forecast period, safety margin lower limit, and model parameter adjustment, such as regression model variable weights. The remote client is compatible with Windows / Linux systems, enabling remote operation and monitoring. During the decision-making process, the system provides real-time visualization of the operating status of each module and key calculation results, such as efficiency ranking, safety margin distribution, and optimization iteration process. After the decision is made, the system supports exporting investment plans to formats such as Excel and PDF, and also receives user adjustment commands to trigger secondary system optimization.

[0070] Investment Forecasting and Analysis Module: Core of Dynamic Forecasting of Total Investment

[0071] Working principle: Based on a multiple linear regression model, the model parameters are trained using the least squares method to construct the prediction formula: I t =α0+∑α it x it +ε

[0072] Among them, I t As a predictor variable, x it The variables collected are α0 and α. it ε represents the influence coefficient, and ε represents the error term. The module incorporates a leave-one-out cross-validation mechanism to evaluate the model's robustness and outputs the predicted total power grid investment value I for the next t+n periods. t+n And transmit it to the investment equilibrium analysis module;

[0073] Principle of the preferred technical solution:

[0074] Model adaptive update: The latest historical data is automatically retrieved every quarter, such as load growth and investment completion status for newly added quarters, and α0 and α are re-optimized iteratively using the least squares method. it Parameters such as these ensure that the model adapts to dynamic changes in the power grid;

[0075] Leave-one-out cross-validation: Individual samples are removed sequentially from the collected sample data, and the remaining samples are used to train the model and predict the values ​​of the removed samples. By calculating the deviation between the predicted values ​​and the actual values, the extrapolation ability and robustness of the model are evaluated, and overfitting is avoided.

[0076] Output: The verified predicted total investment amount I for the next t+n periods. t+n This data is transmitted to the investment equilibrium analysis module as a total constraint.

[0077] The investment efficiency analysis module uses a four-stage DEA model to isolate the interference of environmental factors and random errors on efficiency, and outputs an unbiased efficiency value that reflects the true management level of the region. The specific four-stage logic is as follows:

[0078] Phase 1: Initial efficiency calculation. Based on the BCC model (variable returns to scale), construct the constraint condition minθ. k The original efficiency value θ of each decision unit (DMU) is solved by linear programming. k (Pure technical efficiency) is essentially a measure of relative efficiency at the current level of input and output.

[0079] The DEA model described above is a nonparametric efficiency evaluation method based on linear programming. Its core principle is to determine the effective production frontier by comparing the inputs and outputs of decision-making units, such as the power grids of different regions. Units not falling on the frontier are considered relatively inefficient, and their distance from the frontier represents the efficiency value range of 0-1. This framework does not rely on a specific production function form and can evaluate the efficiency of complex systems with high inputs and high outputs using only actual data, perfectly meeting the evaluation needs of power grid investments that involve "higher capital and equipment inputs and higher power outputs and reliability."

[0080] Using the BCC model in the DEA model, with power grid investment amount and operation and maintenance cost as input indicators, and power supply quantity and power supply reliability as output indicators, the initial pure technical efficiency and comprehensive efficiency of each region are calculated: Comprehensive efficiency = Pure technical efficiency × Scale efficiency.

[0081] Phase Two: Separating Environment and Error, Constructing the SFA Regression Model rk =f(Z) k ;β r )+V rk +U rk

[0082] Among them, S rk f(Z) represents the slack variable of the nth input of the k-th DMU; k ;β r ) represents the effect of external environmental factors on the slack variable S rk The influence, take f(Z) k ;β r ) = Z k·β r Z k β represents observable environmental factor variables. r Z represents the environmental factor variable k The corresponding parameter vector that needs to be estimated; V rk +U rk V represents the mixed error term. rk Let U be the random error term, assumed to follow a normal distribution. rk To manage inefficiency terms, we assume they follow a truncated normal distribution, V rk with U rk Independent and unrelated.

[0083] Environmental variable optimization: Select regional economic level (GDP), power grid scale, such as line length, and climate conditions, such as annual average temperature, as core environmental variables to comprehensively cover the impact of non-management factors on efficiency;

[0084] Error distribution assumption: Assume V rk Follows a normal distribution U rk Follows a truncated normal distribution β is estimated using the maximum likelihood estimation method. r σ 2 And parameters such as γ, and then through the expectation formula Calculate the inefficiency item U in management rk Accurately separate environmental, random errors, and management factors.

[0085] Phase Three: Adjusting Input Data. This is done using formulas. Adjust the input data of all decision-making units to a homogeneous environment and random state to eliminate interference from external factors, where X rk For the actual input of the decision-making unit, To adjust the input, [max(Z) k ·β r )-Z k ·β r ] indicates adjusting different decision-making units to homogeneous environmental conditions, [max(V rk )-V rk This indicates adjusting different decision-making units to the same state of nature;

[0086] Phase 4: Unbiased Efficiency Reassessment. The adjusted input data is substituted into the BCC model to recalculate the pure technical efficiency (PTE), overall efficiency (TE), and scale efficiency (SE), and the unbiased efficiency values ​​for each region are output to the investment equilibrium analysis module.

[0087] Safety margin analysis module: focuses on the quantification of power grid operation safety status and investment gain modeling, and realizes dynamic safety constraint output;

[0088] Safety margin quantification: Define the power grid operation safety domain, such as a reasonable capacity-to-load ratio range of 1.5-2.0 for 220kV and 1.3-1.8 for 110kV, and support custom configuration of constraint parameters according to voltage level; use the analytic hierarchy process (AHP) to determine the weights of indicators such as capacity-to-load ratio, load factor, and power flow distribution, calculate the weighted distance from the current operating point to the safety domain boundary, and obtain the real-time safety margin s. i The greater the distance, the higher the safety level; Investment safety gain modeling: Two modeling methods are provided: regression analysis based on historical data, using investment and safety margin changes over the years, and training the gain function through a random forest algorithm; and power grid digital simulation based on technical principles, simulating the impact of investment projects such as new lines and transformer capacity expansion on power flow and voltage, quantifying the safety margin improvement value, and finally establishing the functional relationship f. i (I i ), I i For regional investment quotas;

[0089] Output: Current safety margin s for each region i and investment safety gain function f i (I i This provides a basis for safety constraints to optimize investment balance.

[0090] Investment Equilibrium Analysis Module: Multi-Objective Collaborative Optimization Core Working Principle: As the core engine of the system, it integrates the output data of preceding modules and outputs the optimal investment allocation scheme through a four-step process: initial allocation, equilibrium calculation, weight determination, and multi-objective optimization. Step 1: Initial Allocation. Based on the previous period's comprehensive efficiency value TE for each region... i,t-1 Through formula Complete the initial allocation of total investment, prioritizing efficiency.

[0091] Step 2: Balance Calculation. Plot the Lorentz curves for the increase in electricity sales per unit investment, the 220kV capacity-to-load ratio, and the 110kV capacity-to-load ratio. Sort the unit index values ​​in ascending order and calculate the Gini coefficient G using the trapezoidal area method. j The formula is

[0092]

[0093] A smaller Ni coefficient indicates better balance. In the formula, j represents the index number for increased electricity sales per unit investment, 220 kV capacity-to-load ratio, and 110 kV capacity-to-load ratio; i represents the unit number; G j X is the Gini coefficient based on a certain index j; j(i) M is the cumulative percentage of indicator j; j(i)Y is the index value of the i-th unit j; j(i) W represents the cumulative percentage of investment allocation based on indicator j. i Let m be the investment amount for the i-th unit; m is the number of allocation units; when i = 1, (X i-1 ,Y i-1 ) is considered as (0,0).

[0094] Step 3: Determine the weights by calculating the weights ω of each equilibrium index using the entropy method. j First, standardize the indicator data, then calculate the information entropy using the following formula:

[0095]

[0096] Finally passed Determine the weights to ensure that the weight allocation is objective and unbiased;

[0097] In the formula, x i Let z represent the initial investment allocation for the i-th unit. ij y represents the j-th index value of the i-th unit. ij p represents the unit investment amount of the j-th indicator in the i-th unit. ij This represents the proportion of the i-th unit under the j-th indicator in that indicator, e j ω represents the information entropy of the unit investment amount for the j-th indicator. j This represents the weight of the j-th indicator, where j = 1, 2, 3.

[0098] Step 4: Multi-objective optimization. Construct an optimization model with the objective of minimizing the weighted sum of the Gini coefficients. The objective function is... The constraints include:

[0099] Total investment adjustment ratio constraint:

[0100] Constraints on the adjustment ratio of investment amount for each unit:

[0101] Dynamic safety constraints: s i +f i (I i )≥s min ;

[0102] Gini coefficient constraints for each indicator: G 1(j) ≤G 0(j) ;

[0103] In the formula, F is the weighted average of the Gini coefficients of each indicator, j is the number of the three indicators, i is the allocation unit number, and W... 0(i) W 1(i)These represent the investment amounts for each unit before and after the adjustment, respectively. q0 and q1 are the upper and lower limits of the total investment adjustment ratio, respectively. i0 and p i1 G sets the upper and lower limits for adjusting the investment amount of each unit. 0(j) Let be the initial value of the Gini coefficient corresponding to index j.

[0104] Principle of the preferred technical solution:

[0105] Algorithm improvement: An improved particle swarm optimization algorithm is adopted to solve the problem, and an inertia weight dynamic adjustment strategy is introduced. The weight range is 0.4-0.9. In the early stage of iteration, the high weight enhances the global search, and in the later stage, the low weight improves local convergence.

[0106] Dynamic adjustment of safety margin constraints: Based on the early warning level of the power grid operation status, such as normal, watch, warning, and emergency, the lower limit of the safety margin (s) is automatically adjusted. min , such as in an emergency min The target was raised from 0.8 to 1.0 to ensure that the investment plan prioritizes security.

[0107] Output: The final power grid investment allocation scheme that satisfies all constraints.

[0108] The dispatch automation system, infrastructure management system, production management system, and SCADA system that the data acquisition unit connects to through the power system's dedicated communication protocol are all mature existing technologies in the power industry. The dispatch automation system is the "nerve center" of the power system's automated decision-making and the "digital management center" of the new power infrastructure. The production management system is the "digital operation center" of the power production process, and the SCADA system is a core branch of industrial control systems (ICS).

[0109] In the above technical solution, the touch screen uses an industrial-grade dedicated touch device, such as the Advantech TPC1571Gi series, and the remote client uses an industrial-grade remote monitoring terminal, such as the Advantech UNO-2484G industrial-grade remote host structure.

[0110] Detailed Workflow

[0111] The system workflow follows a closed-loop logic of data acquisition, analysis and processing, optimization decision-making, and output feedback. The specific steps are as follows:

[0112] Step 1: Data Acquisition and Standardization

[0113] The data acquisition unit connects to data sources such as the dispatch automation system and SCADA system via IEC61850 and DL / T860 protocols, and collects key indicators such as historical investment, load growth and capacity ratio at a frequency of 15 minutes / time.

[0114] The built-in edge computing module preprocesses the raw data: it uses mean filtering to remove high-frequency noise, linear interpolation to fill in missing data, and standardizes the data format to CSV to ensure data interface compatibility between modules.

[0115] The preprocessed data is synchronously transmitted to the data storage unit for backup, and simultaneously distributed to the investment forecasting and analysis module, the investment efficiency analysis module, and the safety margin analysis module.

[0116] Step 2: Total Investment Forecast

[0117] The investment forecasting and analysis module calls a multiple linear regression model and loads standardized historical data, such as investment and load growth and the proportion of new energy access over the past 5 years.

[0118] The model parameters α0 and α are trained using the least squares method. it The robustness of the model is tested using the leave-one-out cross-validation method. If the prediction error is ≤5%, proceed to the next step; otherwise, re-optimize the model variables.

[0119] Input future scenario parameters, such as a prediction period of t+n=1 year and a preset load growth rate of 8%. The model outputs a predicted total power grid investment value I for the next year. t+1 For example, 12 billion yuan is transmitted to the investment balance analysis module and simultaneously stored in the data storage unit.

[0120] Step 3: Unbiased efficiency assessment

[0121] The investment efficiency analysis module loads standardized input-output data, such as input indicators: line investment amount, equipment operation and maintenance costs; output indicators: power supply, power supply reliability, and environmental variable data.

[0122] Phase 1: Calculating the original efficiency value θ for each region based on the BCC model. k , 0.65-0.92;

[0123] Phase Two: Constructing an SFA regression model to separate environmental factors, such as regional GDP, power grid size, and random error, and calculating the management inefficiency term U. rk , and the random error term V rk ;

[0124] The third stage: Standardize the input data to a homogeneous environment and random state by adjusting the formula;

[0125] Phase 4: Substitute into the BCC model to re-evaluate efficiency and output the comprehensive efficiency value TE for each region. i,t-1 For example, values ​​between 0.72 and 0.95 are transmitted to the investment equilibrium analysis module.

[0126] Step 4: Safety Margin Diagnosis

[0127] The safety margin analysis module loads real-time operating data from the SCADA system, such as capacity ratio, load factor, and power flow distribution.

[0128] A custom safety domain model is invoked based on voltage levels of 220kV or 110kV. The weights of the indicators are determined using the analytic hierarchy process (AHP), and the real-time safety margin s for each region is calculated. i (0.65-1.12);

[0129] Based on historical investment and safety margin changes data, the investment safety gain function f is trained using the random forest algorithm. i (I i ) = 0.03I i +0.15(I i (Unit: 100 million yuan)

[0130] Output s i with f i (I i The data is then transferred to the investment equilibrium analysis module and simultaneously stored in the data storage unit.

[0131] Step 5: Multi-objective equilibrium optimization

[0132] The investment equilibrium analysis module receives the total investment I. t+1 Overall efficiency value TE i,t-1 Safety margin s i and gain function f i (I i );

[0133] Initial allocation: according to the formula Preliminary allocations have been completed, with initial quotas ranging from 320 million to 780 million yuan for 20 regions.

[0134] Balance calculation: Plot the Lorentz curve and calculate the Gini coefficient G for the increase in electricity sales per unit investment and the capacity-to-load ratio. j (0.25-0.32);

[0135] Weight determination: The index weight ω is calculated using the entropy method. j For example, 0.35, 0.33, 0.32;

[0136] Multi-objective optimization: Construct an optimization model, solve it using an improved particle swarm optimization algorithm, iterate 50 times until convergence, and output the final investment allocation scheme, such as 300-820 million yuan for each region.

[0137] Step 6: Solution Output and Feedback

[0138] The human-computer interaction unit visually displays the total investment, the allocation of funds in each region, and the matching of efficiency and safety margins, and supports exporting the solution in Excel format;

[0139] The data storage unit synchronously stores optimization model parameters, iteration process data, and the final solution, and supports historical traceability;

[0140] If the user adjusts parameters via touchscreen or remote client, such as modifying the lower safety margin s min The system then returns to steps 1-5 to re-execute the optimization until a satisfactory solution is output.

[0141] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-module integrated dynamic optimization decision-making system for power grid investment scale, characterized in that, It includes a data acquisition unit, an investment forecasting and analysis module, an investment efficiency analysis module, a safety margin analysis module, an investment equilibrium analysis module, a data storage unit, and a human-machine interaction unit. Each component communicates with the internal bus via standardized data interfaces. The data acquisition unit interfaces with the dispatch automation system, infrastructure management system, production management system, and SCADA system through a power system-specific communication protocol, automatically collecting power grid investment-related indicators, efficiency evaluation indicators, and safety operation indicators. The investment forecasting and analysis module communicates with the data acquisition unit. The investment efficiency analysis module communicates with the data acquisition unit and uses a four-stage DEA model for unbiased efficiency evaluation. The safety margin analysis module communicates with the data acquisition unit. The investment equilibrium analysis module, as the core optimization engine, communicates with the investment forecasting and analysis module, the investment efficiency analysis module, and the safety margin analysis module.

2. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The built-in multiple linear regression prediction model is trained using the least squares method, with the following formula: I t =α0+∑α it x it +e Among them, I t As a predictor variable, x it The variables collected are α0 and α. it ε represents the influence coefficient, and ε represents the error term. The module incorporates a leave-one-out cross-validation mechanism to evaluate the model's robustness and outputs the predicted total power grid investment value I for the next t+n periods. t+n The data is then transmitted to the investment equilibrium analysis module.

3. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The investment efficiency analysis module includes the following four stages: Phase 1: Preliminary calculation of the original efficiency value based on the BCC model, with the model constraint being: minθ k Where, θ k x represents the pure technical efficiency value of the k-th DMU; ij y represents the i-th type of input for the j-th project; mj Let r represent the m-th output of the j-th project; r and t represent the number of input and output variables of the project, respectively. Phase Two: Constructing an SFA regression model to separate environmental factors from random error, the formula is: S rk =f(Z) k ;β r )+V rk +U rk , among which, S rk f(Z) represents the slack variable of the nth input of the k-th DMU; k ;β r ) represents the effect of external environmental factors on the slack variable S rk The influence, take f(Z) k ;β r ) = Z k ·β r Z k β represents observable environmental factor variables. r Z represents the environmental factor variable k The corresponding parameter vector that needs to be estimated; V rk +U rk V represents the mixed error term. rk For the random error term, when it follows a normal distribution, U rk To manage inefficiency terms, when they conform to a truncated normal distribution, V rk with U rk Independent and uncorrelated, parameters are estimated using the maximum likelihood estimation method, and the input data are adjusted to homogeneous environments and random states; Then, the maximum likelihood estimation method was used to estimate β. r σ 2 And the γ parameter, then calculate the management inefficiency term U. rk : Finally, the formula was adjusted to estimate β. r σ 2 and γ parameter values ​​and management inefficiency terms U rk The random error term V is calculated based on this. rk By adjusting the input values ​​of different decision-making units, they can be placed under homogeneous environmental and weather conditions: Among them, X rk For the actual input of the decision-making unit, To adjust the input, [max(Z) k ·β r )-Z k ·β r ] indicates adjusting different decision-making units to homogeneous environmental conditions, [max(V rk )-V rk This indicates adjusting different decision-making units to the same state of nature; Phase 3: Import the input values ​​of the decision-making units after the adjustments in Phase 2 into the DEA model in Phase 1 to recalculate the relative R&D efficiency of each decision-making unit; Phase 4: Substitute the adjusted data into the BCC model, recalculate the pure technical efficiency, overall efficiency, and scale efficiency, and output the unbiased efficiency values ​​for each region to the investment equilibrium analysis module.

4. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The safety margin analysis module includes: Safety margin quantification: Define the power grid operation safety domain, calculate the weighted distance from the current operating state point to the boundary of the safety domain, and obtain the real-time safety margin s for each region. i ; Investment safety margin modeling: Based on historical data regression analysis or power grid digital simulation, establish a functional relationship f between investment and safety margin improvement. i (I i ), where I i For regional investment quotas; Output: Current safety margin s for each region i and investment safety gain function f i (I i The data is then transmitted to the investment equilibrium analysis module.

5. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The investment equilibrium analysis module includes: Initial allocation: Investment is allocated based on the comprehensive efficiency value of each region, using the following formula: Among them, I i For the investment allocation of unit i in year t, I t Let t be the total investment amount in year t. The percentage of investment efficiency for unit i in the previous year; Balance calculation: Plot the Lorenz curve and use the trapezoidal area method to calculate the Gini coefficient G of the unit investment increase in electricity sales and capacity ratio. j ; Weight determination: The weights ω of each equilibrium index are calculated using the entropy method. j ; Multi-objective optimization: An optimization model is constructed with the objective of minimizing the weighted sum of the Gini coefficients, and conditions including total investment constraints, regional investment adjustment ratio constraints, and safety margin lower bound constraints. Objective function: Constraints: Total investment adjustment ratio constraint: Constraints on the adjustment ratio of investment amount for each unit: Dynamic safety constraints: s i +f i (I i )≥s min ; Gini coefficient constraints for each indicator: G 1(j) ≤G 0(j) ; In the formula, F is the weighted average of the Gini coefficients of each indicator, j is the number of the three indicators, i is the allocation unit number, and W... 0(i) W 1(i) These represent the investment amounts for each unit before and after the adjustment, respectively. q0 and q1 are the upper and lower limits of the total investment adjustment ratio, respectively. i0 and p i1 G sets the upper and lower limits for adjusting the investment amount of each unit. 0(j) Let j be the initial value of the Gini coefficient corresponding to index j; Output: Final power grid investment allocation plan.

6. A multi-module integrated dynamic optimization method for power grid investment scale, employing the multi-module integrated dynamic optimization decision-making system for power grid investment scale as described in any one of claims 1-5, characterized in that, Includes the following steps: S1. Data Acquisition: The data acquisition unit connects to various data sources in the power system to automatically collect the raw data required for investment forecasting, efficiency assessment, and safety analysis, and performs data cleaning and format standardization processing. S2. Total Investment Forecast: The investment forecasting and analysis module calls a multiple linear regression model, inputs standardized data, and outputs the predicted total investment in the future power grid after passing the leave-one-out cross-validation. t+n ; S3. Unbiased efficiency assessment: The investment efficiency analysis module adopts a three-stage DEA model to sequentially complete the original efficiency calculation, separation of environmental and random errors, adjustment of input data and efficiency reassessment, and output the unbiased efficiency value of each region. S4. Safety Margin Diagnosis: The safety margin analysis module quantifies the current safety margin of each region based on real-time operational data. i Establish the investment safety gain function f i (I i ); S5. Multi-objective equilibrium optimization: The investment equilibrium analysis module receives the output data from the preceding module, first completes the initial investment allocation based on the efficiency value, and then constructs and solves the multi-objective optimization model by using the Lorenz curve, Gini coefficient calculation, and entropy method weight determination, and outputs the final investment allocation scheme. S6. Solution Output and Feedback: The investment solution is displayed and exported through the human-computer interaction unit, and the data storage unit stores the relevant data synchronously; if adjustments are needed, return to S1-S5 to re-execute the optimization.

7. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The power grid investment-related indicators include historical investment amounts, load growth data, and power supply structure parameters. The efficiency evaluation indicators include input-output data and environmental variable data. The safe operation indicators include capacity-to-load ratio, load factor, and power flow distribution data.

8. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 3, characterized in that, The investment forecasting and analysis module supports adaptive model updates, automatically calling the latest historical data to retrain the parameters of the multiple linear regression model every quarter. In the SFA regression model of the investment efficiency analysis module, environmental variables include regional economic level, power grid scale, and climate conditions. The random error term adopts the assumptions of normal distribution and truncated normal distribution to improve the accuracy of error separation. The safety domain model of the safety margin analysis module supports custom configuration. The safety constraint parameters can be adjusted according to different voltage levels of 220kV or 110kV. The weights in the weighted distance calculation are determined by the analytic hierarchy process. The optimization model of the investment equilibrium analysis module is solved using an improved particle swarm optimization algorithm, and a dynamic adjustment strategy for inertia weight is introduced. The safety margin constraint of the investment balance analysis module supports dynamic adjustment, and automatically adjusts the value of smin according to the early warning level of the power grid operation status.

9. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The data storage unit adopts an industrial-grade database to store the collected raw data, intermediate calculation results of each module, optimized model parameters and final investment plan. It supports local data backup and cloud synchronization. The data acquisition unit supports IEC61850 and DL / T860 protocols, with a data sampling frequency of 15 minutes / time. It has a built-in edge computing preprocessing function to perform noise reduction and interpolation processing on the raw data.

10. The multi-module integrated dynamic optimization decision-making system for power grid investment scale according to claim 1, characterized in that, The human-computer interaction unit includes a touch screen and a remote client, which supports investment parameter configuration, model parameter adjustment, visualization of the decision-making process, and export of investment plans.

Images