A source-grid-load-storage collaborative planning method and system for a new power system

Abstract

This invention discloses a source-grid-load-storage coordinated planning method and system for new power systems, belonging to the field of distribution network planning technology. This invention addresses the problems in existing planning methods where the strong randomness and volatility of new energy output, along with the increasingly complex coupling relationships among multiple links in the source-grid-load-storage system, make it difficult for traditional deterministic and independent planning models to accurately depict the system's operating state, leading to resource misallocation and investment waste. This invention constructs a multi-scale feature dataset and a mechanistic model describing the coordinated relationships among source-grid-load-storage, generating a digital twin model through hierarchical parameter calibration. It then performs multi-objective coordinated planning calculations and multi-dimensional evaluations under multiple planning scenarios, thereby achieving high-precision and highly adaptable source-grid-load-storage coordinated optimization planning for new power systems.

Description

Technical Field

[0001] This invention relates to the field of power distribution network planning technology, specifically to a source-grid-load-storage coordinated planning method and system for new power systems. Background Technology

[0002] With the deepening of the "dual carbon" goals, a new type of power system, mainly based on new energy sources such as wind and solar, is being rapidly constructed. Compared with the traditional power system, the new power system exhibits significant characteristics such as a high proportion of renewable energy integration, high power electronics, enhanced load-side interaction capabilities, and widespread application of energy storage. This change presents new challenges to power system planning: on the one hand, the strong randomness, volatility, and intermittency of new energy output make it difficult for traditional deterministic planning methods to accurately depict the system's operating state; on the other hand, the coupling relationships between the sources (new energy power generation), grid (transmission and distribution network), load (adjustable load), and storage (energy storage) are becoming increasingly complex, and traditional independent planning models are prone to resource misallocation and investment waste.

[0003] Existing planning methodologies often focus on optimizing single aspects (such as considering only power supply or grid expansion) or rely on simplified typical scenarios or empirical coefficients, making it difficult to fully uncover the hidden patterns in massive operational data. Furthermore, the lack of effective integration between physical mechanism models and data-driven models results in insufficient accuracy in the dynamic mapping of the planning model to the actual system. In addition, source-load uncertainties are often handled with fixed margins or conservative scenario sets during the planning phase, lacking systematic evaluation methods for multi-objective trade-offs under various planning scenarios.

[0004] Therefore, there is an urgent need for a planning method that can integrate big data analysis with physical mechanisms, achieve coordinated optimization of power generation, grid, load and storage, and support multi-dimensional evaluation, so as to improve the economy, reliability and adaptability of new power system planning. To this end, we propose a coordinated planning method and system for power generation, grid, load and storage for new power systems. Summary of the Invention

[0005] The purpose of this invention is to provide a source-grid-load-storage collaborative planning method and system for novel power systems. It constructs a multi-scale feature dataset through big data analysis and mining, and establishes a mechanistic model describing the coupling relationship between sources, grid, load, and storage based on the physical mechanisms of the power system. Then, it uses the multi-scale feature dataset to perform hierarchical parameter calibration on the mechanistic model, generating a digital twin model that is synchronously mapped to the physical system in real time. Multiple planning scenarios are generated in this model and associated with uncertainty sets. Multi-objective collaborative planning calculations are performed to obtain a set of candidate solutions. Finally, the optimal solution is selected through multi-dimensional evaluation, thus solving the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A source-grid-load-storage coordinated planning method for new power systems includes the following steps: Acquire historical operating data of new power systems, perform big data analysis and mining, and construct a multi-scale feature dataset of power sources, grids, loads, and storage. Based on the physical mechanisms of power systems, a mechanism model describing the collaborative relationship between power generation, grid, load, and storage is constructed. The mechanism model is calibrated using the multi-scale feature dataset to generate a digital twin model that is synchronously mapped to the physical power system in real time. Multiple planning scenarios are generated in the digital twin model, and a set of uncertainties is associated with each planning scenario; For each planning scenario, multi-objective collaborative planning calculations are performed in the digital twin model to generate a set of candidate planning schemes; The results of big data analysis and mining are used to evaluate the candidate planning scheme set from multiple dimensions, and the optimal collaborative planning scheme is selected.

[0007] Furthermore, big data analysis and mining include: A hierarchical clustering algorithm based on dynamic time warping is used to divide the minute-level load data and the minute-level output data of each new energy power station into time-series patterns, and extract the typical daily load curve, typical daily output curve and probability distribution corresponding to each typical scenario. A feature selection method based on mutual information is adopted, with the new energy output sequence as the target variable and the multi-source meteorological data sequence as the feature variable. The mutual information value between each meteorological factor and the output sequence is calculated, and the dominant meteorological factors are selected by sorting the mutual information values. The transfer entropy method is used to identify the lag correlation between the dominant meteorological factors and the output sequence. The extracted typical scene curves, probability distributions, and identified dominant meteorological factors and their lag correlations are fused and encoded to construct a multi-scale feature dataset of source-grid-load-storage that includes time-domain features, frequency-domain features, and probability distribution features.

[0008] Furthermore, the mechanistic model includes four interconnected and coupled sub-mechanistic models, specifically: The new energy mechanism sub-model is used to describe the active-reactive coupling output characteristics of new energy units. It embeds the irradiance-temperature-power conversion function and the maximum power point tracking control logic. It is used to drive the meteorological input data of new energy power plants and output the time-series predicted values ​​of active power and reactive power. The power flow mechanism sub-model is used to describe the power transmission and voltage distribution characteristics of the power transmission and distribution network. It adopts the node power balance equation in polar coordinate form and introduces the adjustment logic of the on-load tap changer of the transformer and the automatic switching logic of the parallel capacitor bank. It takes the injected power of each node as input and outputs the voltage amplitude and phase angle of each node, and the active power and reactive power of each branch. The load response mechanism sub-model is used to describe the demand response characteristics of adjustable loads. It includes a load elasticity coefficient matrix based on real-time electricity prices and load interruption and load transfer contract logic based on demand-side response contracts. It takes electricity price signals and control instructions as inputs and outputs the actual response power of load nodes. The energy storage mechanism sub-model is used to describe the charging and discharging dynamic characteristics of energy storage devices, including the charging and discharging efficiency coefficient, self-discharge rate and cycle life decay factor. It takes the charging and discharging power command as input and outputs the recursive value of the state of charge time series and the available capacity. The four mechanism sub-models are linked by coupling variables, which include at least: the voltage amplitude and phase angle of each node, the power of each branch, and the injected power of each node, forming a hybrid driving mechanism model in the form of a hybrid integer nonlinear programming.

[0009] Furthermore, the stratification parameter calibration includes: First layer: Using historical power output data from renewable energy power plants, identify the power conversion coefficient and time delay parameters in the renewable energy power output mechanism sub-model online; The second layer: using historical voltage and power data of each node in the distribution network, online estimation of line impedance parameters and transformer turns ratio parameters in the power flow mechanism sub-model; The third layer: After locking the identification results of the first and second layers, the charging and discharging efficiency and capacity decay coefficient in the energy storage mechanism sub-model are identified online using the historical operating data of the energy storage system. The parameters after hierarchical calibration are substituted into the mechanistic model to generate the digital twin model, which has a model error feedback interface for rolling correction.

[0010] Furthermore, multiple planning scenarios are generated in the digital twin model, and a set of uncertainties is associated with each planning scenario, specifically including: Simulate a scenario of increased new energy penetration by modifying installed capacity parameters; By modifying the load baseline curve and overlaying the load morphology variation pattern extracted from a multi-scale feature dataset, a load growth scenario is simulated. Different energy storage configuration schemes can be simulated by modifying the capacity and power parameters of the energy storage device at a specified node. Based on the uncertainty intervals obtained from big data analysis and mining, each planning scenario is associated with a set of uncertainties represented in the form of interval numbers or fuzzy numbers.

[0011] Furthermore, multi-objective collaborative programming calculations include: The system aims to minimize the total life-cycle cost of the system and maximize the annual consumption of renewable energy as the synergistic optimization objectives, and uses nodal power balance constraints, line power flow limit constraints, new energy output uncertainty constraints, energy storage state of charge constraints, and adjustable load response capacity constraints as operational constraints. The above planning model is solved by a multi-objective evolutionary algorithm, resulting in a set of candidate planning schemes. Any scheme in this set of candidate planning schemes cannot further optimize any objective without degrading the other objective. During the solution process, the model error feedback interface of the digital twin model is used to receive measured data for rolling parameter correction.

[0012] Furthermore, the multi-dimensional assessment includes: The total lifecycle cost of each candidate planning scheme is used as the first dimension of evaluation value; In the digital twin model corresponding to each of the candidate planning schemes, N-1 fault scans are performed one by one, and the resulting power shortage probability and expected power shortage amount are used as the second dimension evaluation value. For each candidate planning scheme associated with an uncertainty set, the extreme value difference between the two optimization objectives under each candidate planning scheme within the range of the uncertainty set is output, and the extreme value difference is used as the third dimension evaluation value. The normalized values ​​of each candidate planning scheme in the first, second, and third dimensions are used as inputs to the fuzzy hierarchical analysis method, and the subjective weights of each dimension are output. The normalized values ​​are used as input to the entropy weight method, and the objective weights of each dimension are output. The subjective weights and objective weights are multiplied and combined to output the comprehensive weight; Based on the comprehensive weight, each candidate planning scheme is weighted and ranked, and the candidate planning scheme with the highest ranking is output as the optimal collaborative planning scheme.

[0013] A source-grid-load-storage coordinated planning system for a new type of power system is configured to execute a source-grid-load-storage coordinated planning method for a new type of power system, including: The multi-scale feature extraction module is configured to be connected to the data acquisition and monitoring control system of the new power system, acquire multi-year continuous operation data, run big data analysis and mining process, and output multi-scale feature datasets of source, grid, load and storage. The mechanism model construction module is configured to call a pre-set power system component parameter library to establish mathematical equations for new energy generating units, power transmission and distribution networks, adjustable loads and energy storage devices, and to form a mechanism model in the form of mixed integer nonlinear programming by coupling variables. The twin mapping module is configured to receive the multi-scale feature dataset to perform hierarchical parameter identification on the mechanism model, establish an error feedback channel between the identification results and the real-time measurement values ​​of the physical system, and output a dynamically corrected digital twin model. The scenario construction module is configured to inject adjustments to the penetration rate of new energy sources, the load growth rate, and modifications to the energy storage configuration into the digital twin model in the form of parameter perturbation, generate multiple planning scenarios, and add interval constraints to each planning scenario based on the uncertainty boundary obtained from big data analysis and mining. The collaborative optimization solution module is configured to perform a multi-objective evolutionary search under each planning scenario, with a specified bi-objective function as the optimization direction and the operational constraint boundary provided by the digital twin model as the feasible region, to generate a set of candidate planning schemes. The solution decision module is configured to receive the set of candidate planning solutions, call historical statistical features obtained from big data analysis and mining to construct a multi-dimensional evaluation matrix, and output the optimal collaborative planning solution after comprehensive evaluation and screening.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention constructs a digital twin model that is synchronously mapped to the physical power system in real time by integrating big data analysis and physical mechanism modeling. This overcomes the shortcomings of traditional deterministic planning methods in accurately depicting the strong randomness, volatility and intermittency of new energy output, thereby improving the mapping accuracy of the planning model to the dynamic operating state of the actual system.

[0015] 2. This invention breaks through the limitations of traditional single-link optimization or independent planning modes, which are prone to resource misallocation and investment waste, by establishing a mechanism sub-model of source-grid-load-storage linkage and multi-objective collaborative planning calculation. Thus, it realizes global collaborative optimization of power source, grid, load and energy storage in the planning stage.

[0016] 3. This invention generates multiple planning scenarios and associates them with uncertainty sets in a digital twin model, replacing the traditional method of handling uncertainty with fixed margins or conservative scenario sets. It combines multi-dimensional evaluation to select the optimal solution, effectively improving the economy, reliability and adaptability of new power system planning. Attached Figure Description

[0017] Figure 1 This is a flowchart of the source-grid-load-storage coordinated planning method for novel power systems according to the present invention; Figure 2 This is an execution diagram of the source-grid-load-storage coordinated planning method for novel power systems according to the present invention; Figure 3 This is a block diagram of the source-grid-load-storage coordinated planning system for a new type of power system according to the present invention. Detailed Implementation

[0018] 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.

[0019] Please see Figure 1-2 The present invention provides a source-grid-load-storage coordinated planning method for a new type of power system, comprising the following steps: Step 1: Obtain historical operating data of the new power system, perform big data analysis and mining, and construct a multi-scale feature dataset of power generation, grid, load, and storage. In one example, the big data analysis and mining includes using a hierarchical clustering algorithm based on dynamic time warping to divide minute-level load data and new energy output data into time-series patterns, and extracting typical daily curves and their probability distributions. In another example, it also includes using a feature selection method based on mutual information to screen dominant meteorological factors and identify their lag correlation with the output sequence. It can be understood that the multi-scale feature dataset integrates time-domain features, frequency-domain features, and probability distribution features. Step 2: Based on the physical mechanisms of the power system, construct a mechanism model describing the collaborative relationship between power generation, grid, load, and energy storage. In one example, the mechanism model includes interconnected sub-models of new energy mechanisms, power flow mechanisms, load response mechanisms, and energy storage mechanisms. It can be understood that the four mechanism sub-models are interconnected through coupling variables such as node voltage, branch power, and node injected power, forming a hybrid driven mechanism model in the form of mixed integer nonlinear programming. Step 3: Perform hierarchical parameter calibration on the mechanistic model using the multi-scale feature dataset to generate a digital twin model that is synchronously mapped to the physical power system in real time. In one example, the hierarchical parameter calibration can be divided into three levels. After substituting the hierarchically calibrated parameters into the mechanistic model, a model error feedback interface for rolling correction needs to be established. This interface receives the measured values ​​of the physical system in real time and compares them with the model output, feeding the error signal back into the model to achieve online parameter correction. It can be understood that after the calibrated parameters are substituted into the mechanistic model, a digital twin model with a model error feedback interface is generated for subsequent rolling correction. Step 4: Generate multiple planning scenarios in the digital twin model and associate an uncertainty set with each planning scenario; in one example, simulate the scenario of increasing new energy penetration rate by modifying the installed capacity parameters, and simulate the scenario of load growth by modifying the load baseline curve and superimposing the load pattern variation mode; in another example, simulate different energy storage configuration schemes by modifying the capacity and power parameters of energy storage equipment at a specified node; it can be understood that, based on the uncertainty interval obtained by big data analysis and mining, an uncertainty set represented in the form of interval number or fuzzy number is associated with each planning scenario. Step 5: For each planning scenario, perform multi-objective collaborative planning calculations in the digital twin model to generate a set of candidate planning schemes. In one example, the collaborative optimization objectives are minimizing the system's total life-cycle cost and maximizing the annual renewable energy consumption, with operational constraints including node power balance constraints, line power flow limit constraints, new energy output uncertainty constraints, energy storage state of charge constraints, and adjustable load response capacity constraints. In another example, a multi-objective evolutionary algorithm is used to solve the above planning model to obtain a set of optimal candidate planning schemes. It is understood that during the solution process, the model error feedback interface of the digital twin model is used to receive measured data for rolling parameter correction. Step 6: Utilize the results of big data analysis and mining to conduct a multi-dimensional evaluation of the candidate planning scheme set, and select the optimal collaborative planning scheme. In one example, the multi-dimensional evaluation includes three dimensions. After normalizing the evaluation values ​​of the three dimensions, the subjective weights are obtained by inputting fuzzy hierarchical analysis and the objective weights by inputting entropy weighting. The subjective weights and objective weights are then multiplied and combined to obtain the comprehensive weight. In one example, the candidate planning schemes are weighted and ranked based on the comprehensive weight, and the candidate planning scheme with the highest ranking is output as the optimal collaborative planning scheme. Those skilled in the art will understand that the "screening" in this invention refers to the weighted ranking and selection based on multi-dimensional comprehensive evaluation, rather than a simple comparison of a single indicator. For example, if candidate scheme A is better than scheme B in the first dimension but worse than scheme B in the second dimension, a comprehensive judgment is made by combining subjective and objective weights. The final output optimal scheme is the one that achieves a balance between economy, reliability, and robustness.

[0020] In the above embodiments, big data analysis and mining include: A hierarchical clustering algorithm based on dynamic time warping is used to divide the minute-level load data and the minute-level output data of each new energy power station into time-series patterns, and extract the typical daily load curve, typical daily output curve and probability distribution corresponding to each typical scenario. A feature selection method based on mutual information is adopted, with the new energy output sequence as the target variable and the multi-source meteorological data sequence as the feature variable. The mutual information value between each meteorological factor and the output sequence is calculated, and the dominant meteorological factors are selected by sorting the mutual information values. The transfer entropy method is used to identify the lag correlation between the dominant meteorological factors and the output sequence. The extracted typical scene curves, probability distributions, and identified dominant meteorological factors and their lag correlations are fused and encoded to construct a multi-scale feature dataset of source-grid-load-storage that includes time-domain features, frequency-domain features, and probability distribution features.

[0021] In a specific example, hierarchical clustering based on dynamic time warping, compared to traditional Euclidean distance clustering, can effectively align the scaling and deformation of load or output curves on the time axis, avoiding clustering distortion caused by phase shifts due to changes in wind speed and illumination. The mutual information value ranges from [0, +∞), with a larger value indicating a stronger correlation between the meteorological factor and the output sequence. Compared to traditional mutual information, transfer entropy can distinguish causal directions and identify the lag relationship where meteorological changes precede output fluctuations, such as identifying specific lag times like irradiance changes leading photovoltaic output by 15 minutes and wind speed changes leading wind power output by 10 minutes.

[0022] The technical effects of the above-mentioned technical solution are as follows: Through the above-mentioned big data analysis and mining, on the one hand, statistically significant typical scenarios and their probability distributions are extracted from massive historical operation data, which can provide a reasonable scenario set basis for subsequent multi-scenario planning. On the other hand, through mutual information filtering and transfer entropy causal analysis, the key meteorological factors affecting the output of new energy and their lead time can be quantitatively identified, thereby effectively supporting the characterization of the uncertainty boundary of new energy output and the construction of the input features of the prediction model. The final multi-scale feature dataset integrates time domain morphology, frequency domain components and probability distribution information, which can overcome the limitations of traditional planning methods that rely only on empirical assumptions or single statistical parameters, thereby improving the accuracy of the digital twin model in representing the dynamic characteristics of the actual system.

[0023] In the above embodiments, the mechanism model includes four interconnected and coupled mechanism sub-models, specifically: The new energy mechanism sub-model is used to describe the active-reactive coupling output characteristics of new energy units. It embeds the irradiance-temperature-power conversion function and the maximum power point tracking control logic. It is used to drive the meteorological input data of new energy power plants and output the time-series predicted values ​​of active power and reactive power. The power flow mechanism sub-model is used to describe the power transmission and voltage distribution characteristics of the power transmission and distribution network. It adopts the node power balance equation in polar coordinate form and introduces the adjustment logic of the on-load tap changer of the transformer and the automatic switching logic of the parallel capacitor bank. It takes the injected power of each node as input and outputs the voltage amplitude and phase angle of each node, and the active power and reactive power of each branch. The load response mechanism sub-model is used to describe the demand response characteristics of adjustable loads. It includes a load elasticity coefficient matrix based on real-time electricity prices and load interruption and load transfer contract logic based on demand-side response contracts. It takes electricity price signals and control instructions as inputs and outputs the actual response power of load nodes. The energy storage mechanism sub-model is used to describe the charging and discharging dynamic characteristics of energy storage devices, including the charging and discharging efficiency coefficient, self-discharge rate and cycle life decay factor. It takes the charging and discharging power command as input and outputs the recursive value of the state of charge time series and the available capacity. The four mechanism sub-models are linked by coupling variables, which include at least: the voltage amplitude and phase angle of each node, the power of each branch, and the injected power of each node, forming a hybrid driving mechanism model in the form of a hybrid integer nonlinear programming.

[0024] In a specific example, the maximum power point tracking control logic in the new energy mechanism sub-model is simulated using the perturbation-observation method, and the irradiance-temperature-power conversion function is constructed based on the single-diode equivalent circuit model of photovoltaic modules or the aerodynamic power curve of wind turbines. The on-load tap changer regulation logic in the power flow mechanism sub-model uses discrete tap modeling, and the automatic switching logic of parallel capacitor banks is triggered based on node voltage over-limit thresholds. The resilience coefficient matrix in the load response mechanism sub-model is a diagonally dominant matrix, where diagonal elements represent the sensitivity of each node's load to its own node's electricity price, and off-diagonal elements represent the cross-influence of electricity prices between different nodes. The cycle life decay factor in the energy storage mechanism sub-model uses the rainflow counting method to extract the mapping relationship between charge / discharge depth and cycle count.

[0025] The technical effects of the above-mentioned technical solution are as follows: Through the linkage and coupling of the above four sub-models, the mechanism model realizes the unified mathematical representation and coordinated interaction of four types of heterogeneous components, namely, source, grid, load and storage, in the planning stage. Among them, the new energy mechanism sub-model makes up for the deficiency of treating new energy as a constant power source in traditional planning. The power flow mechanism sub-model introduces the refined action logic of discrete control equipment. The load response mechanism sub-model quantifies the flexible resources on the demand side as controllable variables rather than fixed loads. The energy storage mechanism sub-model takes into account the nonlinear impact of lifetime decay on available capacity. The above four sub-models form a closed-loop feedback through bidirectional interaction of coupling variables such as node voltage, branch power and node injected power. This enables the mechanism model to accurately reflect the bidirectional coupling characteristics in the new power system, thereby providing an interpretable and solvable model foundation for subsequent digital twin mapping.

[0026] In the above embodiments, the hierarchical parameter calibration includes: First layer: Using historical power output data from renewable energy power plants, identify the power conversion coefficient and time delay parameters in the renewable energy power output mechanism sub-model online; The second layer: using historical voltage and power data of each node in the distribution network, online estimation of line impedance parameters and transformer turns ratio parameters in the power flow mechanism sub-model; The third layer: After locking the identification results of the first and second layers, the charging and discharging efficiency and capacity decay coefficient in the energy storage mechanism sub-model are identified online using the historical operating data of the energy storage system. The parameters after hierarchical calibration are substituted into the mechanistic model to generate the digital twin model, which has a model error feedback interface for rolling correction.

[0027] In a specific example, the power conversion coefficients in the first layer include the photovoltaic conversion efficiency coefficient and temperature correction coefficient of the photovoltaic module, the wind energy capture coefficient and pitch angle control gain coefficient of the wind turbine. The time delay parameter uses cross-correlation analysis to identify the lag time between meteorological changes and power output response, with a value range of 0–30 minutes. The line impedance parameters in the second layer are estimated online using a regularized recursive least squares algorithm, and the transformer turns ratio parameters are identified by inversion based on historical tap changer operation records. The charging and discharging efficiency identification in the third layer combines the Coulomb counting method and the open-circuit voltage method, and the capacity decay coefficient is fitted using an exponential decay model based on the number of cycles. During the layered calibration process, the parameters identified in each layer serve as prior constraints for the calibration of the next layer, avoiding overfitting and unidentifiable problems caused by joint identification of multiple parameters. The model error feedback interface uses the residual between the measured value and the model prediction value as input, and uses Kalman filtering or recursive least squares algorithms to continuously update key parameters, with an update cycle of 5–15 minutes.

[0028] The technical effects of the above solution are as follows: Firstly, the new energy output parameters and grid topology parameters are decoupled and identified through the hierarchical parameter calibration architecture, avoiding the ill-conditioned matrix problem caused by parameter coupling in traditional joint identification. Secondly, the parameters of the new energy and grid layers are locked before calibrating the energy storage parameters, effectively utilizing the characteristics of known charging and discharging power commands and recursive state of charge of the energy storage system, thereby improving the identification accuracy of the capacity decay coefficient. The digital twin model generated after hierarchical calibration not only possesses static parameters that are one-to-one mapped to the physical system, but also achieves online rolling correction of model parameters through an error feedback interface, enabling the digital twin model to continuously track the slow time-varying characteristics of the physical system, such as aging, wear, and seasonal changes. Compared with traditional offline calibration and fixed-use modeling methods, the digital twin model generated by this invention can maintain high dynamic mapping accuracy throughout its entire lifecycle, thus providing a reliable foundation for subsequent multi-scenario planning and multi-objective optimization.

[0029] In the above embodiments, multiple planning scenarios are generated in the digital twin model, and an uncertainty set is associated with each planning scenario, specifically including: Simulate a scenario of increased new energy penetration by modifying installed capacity parameters; By modifying the load baseline curve and overlaying the load morphology variation pattern extracted from a multi-scale feature dataset, a load growth scenario is simulated. Different energy storage configuration schemes can be simulated by modifying the capacity and power parameters of the energy storage device at a specified node. Based on the uncertainty intervals obtained from big data analysis and mining, each planning scenario is associated with a set of uncertainties represented in the form of interval numbers or fuzzy numbers.

[0030] The technical effects of the above-mentioned solution are as follows: By flexibly modifying the installed capacity, load baseline curve, and energy storage configuration parameters in the digital twin model, it is possible to systematically generate multiple planning scenarios covering the increase in new energy penetration, load growth, and different energy storage schemes. This avoids the repetitive work of building independent models one by one in traditional methods. At the same time, based on the uncertainty interval obtained from big data analysis and mining, a set of source-load uncertainties represented by interval numbers or fuzzy numbers is associated for each planning scenario. This overcomes the shortcomings of traditional fixed margins or single scenario sets that cannot reflect the actual fluctuation range. The above-mentioned scenario generation and uncertainty association mechanism enables subsequent multi-objective collaborative planning to systematically evaluate the collaborative potential of each link of source, grid, load, and storage under different boundary conditions, thereby improving the adaptability of the planning scheme to the randomness and fuzziness in the real operating environment.

[0031] In the above embodiments, multi-objective collaborative planning calculation includes: The system aims to minimize the total life-cycle cost of the system and maximize the annual consumption of renewable energy as the synergistic optimization objectives, and uses nodal power balance constraints, line power flow limit constraints, new energy output uncertainty constraints, energy storage state of charge constraints, and adjustable load response capacity constraints as operational constraints. The above planning model is solved by a multi-objective evolutionary algorithm, resulting in a set of candidate planning schemes. Any scheme in this set of candidate planning schemes cannot further optimize any objective without degrading the other objective. During the solution process, the model error feedback interface of the digital twin model is used to receive measured data for rolling parameter correction.

[0032] The technical effects of the above-mentioned solution are as follows: By minimizing the system's total life cycle cost and maximizing the annual renewable energy consumption as dual optimization objectives, and by introducing multiple operational constraints such as node power balance, line power flow limits, uncertainty of renewable energy output, energy storage state of charge, and adjustable load response capacity, the planning model can simultaneously consider both economic efficiency and environmental friendliness. This avoids the one-sided decision-making of traditional single-objective optimization. The candidate solution set obtained by the multi-objective evolutionary algorithm can provide decision-makers with multiple alternative solutions that are at the optimal trade-off boundary between economic efficiency and consumption capacity. More importantly, during the solution process, the model error feedback interface of the digital twin model is used to receive measured data in real time to continuously correct the parameters, enabling the planning calculation to dynamically respond to changes in the actual operating state of the physical system, thereby improving the executability of the optimization results in the real operating environment.

[0033] In the above embodiments, the multi-dimensional evaluation includes: The total lifecycle cost of each candidate planning scheme is used as the first dimension of evaluation value; In the digital twin model corresponding to each of the candidate planning schemes, N-1 fault scans are performed one by one, and the resulting power shortage probability and expected power shortage amount are used as the second dimension evaluation value. For each candidate planning scheme associated with an uncertainty set, the extreme value difference between the two optimization objectives under each candidate planning scheme within the range of the uncertainty set is output, and the extreme value difference is used as the third dimension evaluation value. The normalized values ​​of each candidate planning scheme in the first, second, and third dimensions are used as inputs to the fuzzy hierarchical analysis method, and the subjective weights of each dimension are output. The normalized values ​​are used as input to the entropy weight method, and the objective weights of each dimension are output. The subjective weights and objective weights are multiplied and combined to output the comprehensive weight; Based on the comprehensive weight, each candidate planning scheme is weighted and ranked, and the candidate planning scheme with the highest ranking is output as the optimal collaborative planning scheme.

[0034] In a specific example, the first dimension, total lifecycle cost, includes investment cost, operation and maintenance cost, fuel cost, and decommissioning cost, expressed in RMB 10,000. A smaller value indicates better economic efficiency. In the second dimension, the probability of insufficient power supply ranges from [0,1], representing the probability that the system cannot meet load demand under N-1 fault conditions. The expected power shortage is expressed in MWh / year. Both dimensions reflect system reliability, and a smaller value indicates better performance. In the third dimension, extreme value difference refers to the difference between the maximum and minimum values ​​of the two optimization objectives (total lifecycle cost and renewable energy consumption) when the source and load parameters within the associated uncertainty set are traversed to the upper and lower boundaries under the same candidate scheme. A larger difference indicates that the scheme is more sensitive to uncertainty and less robust; therefore, a smaller extreme value difference indicates better performance. Normalization uses the range normalization method to map the evaluation values ​​of each dimension to the [0,1] interval. The fuzzy hierarchical analysis method uses triangular fuzzy numbers to construct the judgment matrix. The entropy weight method determines the objective weights based on the information entropy of each scheme in each dimension. The multiplicative combination formula is: comprehensive weight = subjective weight × objective weight / sum of the products of each dimension. During weighted ranking, the comprehensive weight is used as the coefficient to calculate the weighted score of each scheme, and the scheme with the highest score is output as the optimal collaborative planning scheme.

[0035] The technical effects of the above-mentioned solution are as follows: By constructing a three-dimensional evaluation system covering the entire life cycle cost, the probability of insufficient power supply and the expected power shortage obtained from N-1 fault scanning, and the difference in the extreme value of the optimization target under uncertainty set, the system simultaneously considers the three mutually constraining dimensions of economy, reliability, and robustness in the selection of planning schemes. The normalized evaluation values ​​of the three dimensions are simultaneously input into the fuzzy hierarchical analysis method to obtain subjective weights and into the entropy weight method to obtain objective weights. Then, the comprehensive weight is obtained by multiplication and combination. This enables the organic integration of expert experience and data-driven information, thereby avoiding the subjective bias or objective distortion of a single weighting method. Finally, based on the weighted ranking of the comprehensive weights, the output optimal collaborative planning scheme is a scheme that achieves a comprehensive optimal balance among economy, reliability, and robustness, rather than an extreme scheme that simply emphasizes one dimension, thereby improving the scientific nature of the planning and decision-making of new power systems.

[0036] Please see Figure 3 This invention also provides a source-grid-load-storage coordinated planning system for a new type of power system, configured to execute a source-grid-load-storage coordinated planning method for a new type of power system, including: The multi-scale feature extraction module is configured to be connected to the data acquisition and monitoring control system of the new power system, acquire multi-year continuous operation data, run big data analysis and mining process, and output multi-scale feature datasets of source, grid, load and storage. The mechanism model construction module is configured to call a pre-set power system component parameter library to establish mathematical equations for new energy generating units, power transmission and distribution networks, adjustable loads and energy storage devices, and to form a mechanism model in the form of mixed integer nonlinear programming by coupling variables. The twin mapping module is connected to the multi-scale feature extraction module and the mechanism model construction module respectively. It is configured to receive the multi-scale feature dataset to identify the hierarchical parameters of the mechanism model, establish an error feedback channel between the identification results and the real-time measurement values ​​of the physical system, and output a dynamically corrected digital twin model. The scenario construction module, connected to the twin mapping module, is configured to inject new energy penetration rate adjustment, load growth rate and energy storage configuration modification into the digital twin model in the form of parameter perturbation, generate multiple planning scenarios, and add interval constraints to each planning scenario based on the uncertainty boundary obtained by big data analysis and mining. The collaborative optimization solution module is connected to the scenario construction module and is configured to perform a multi-objective evolutionary search under each planning scenario, with the specified bi-objective function as the optimization direction and the operational constraint boundary provided by the digital twin model as the feasible region, to generate a set of candidate planning schemes. The scheme decision module, connected to the collaborative optimization solution module, is configured to receive the set of candidate planning schemes, construct a multi-dimensional evaluation matrix by calling historical statistical features obtained from big data analysis and mining, and output the optimal collaborative planning scheme after comprehensive evaluation and screening.

[0037] The technical effects of the above solution are as follows: Through the synergy of the multi-scale feature extraction module and the mechanism model construction module, a deep integration of data-driven and physical mechanism approaches can be achieved, thus overcoming the limitations of experience-driven or single-model-driven approaches in traditional planning; the twin mapping module, through hierarchical parameter identification and error feedback channels, can generate a digital twin model that can be rolled over, enabling the system to continuously track the time-varying characteristics of the physical system, thereby improving the accuracy of the model's dynamic mapping of the actual system; the scenario construction module supports the flexible generation of multiple planning scenarios and the addition of interval constraints in the digital twin model using parameter perturbation, thus avoiding the need to build independent scenarios one by one. The system avoids repetitive work in the scenario; the collaborative optimization solution module performs multi-objective evolutionary search under dual objective functions and multiple operational constraints, and the output candidate solution set forms an optimal trade-off between economic efficiency and renewable energy consumption; the solution decision module constructs a three-dimensional evaluation matrix and integrates subjective and objective weighting methods to achieve a comprehensive balance between economic efficiency, reliability, and robustness in the final output optimal solution; the entire system forms a complete closed loop from data acquisition, feature extraction, twin mapping scenario construction, collaborative optimization to solution decision-making, thus providing a solution with dynamic adaptive capability and multi-objective comprehensive trade-off capability for the coordinated planning of power generation, grid, load, and storage in new power systems.

[0038] This invention provides a source-grid-load-storage collaborative planning method and system for novel power systems, applied to a distribution network planning model. Its internal multi-scale feature extraction module extracts multi-scale features of sources, grids, loads, and storage through big data analysis and mining. Then, a mechanism model construction module constructs four sub-models based on physical mechanisms, linked by coupled variables to form a mixed-integer nonlinear programming model. The mechanism model is then calibrated using a multi-scale feature dataset to generate a digital twin model that is synchronously mapped to the physical system in real time, and includes an error feedback interface. Multiple planning scenarios are generated within this digital twin model, along with an uncertainty set. With the dual objectives of minimizing life-cycle cost and maximizing renewable energy consumption, a multi-objective evolutionary search is performed to obtain a set of candidate solutions. Finally, the candidate solutions are comprehensively evaluated in terms of economy, reliability, and robustness through a combination of fuzzy hierarchical analysis and entropy weighting in the solution decision module, thereby outputting the optimal solution. This invention addresses the problems of traditional planning methods, which suffer from difficulty in characterizing system operation, resource misallocation, and investment waste due to the strong randomness and volatility of renewable energy output and the increasingly complex coupling relationships among multiple links in the power generation, grid, load, and storage systems. It significantly improves the economic efficiency and power supply reliability of planning schemes. Whether facing scenarios with a high proportion of renewable energy integration or extreme power output fluctuations, it achieves high-precision and highly adaptable source-grid-load-storage collaborative optimization planning through real-time synchronous mapping of digital twin models and synergy of multi-objective evolutionary search, making the planning results more in line with the actual operational needs of new power systems.

[0039] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A source-grid-load-storage coordinated planning method for new power systems, characterized in that, Includes the following steps: Acquire historical operating data of new power systems, perform big data analysis and mining, and construct a multi-scale feature dataset of power sources, grids, loads, and storage. Based on the physical mechanisms of power systems, a mechanism model describing the collaborative relationship between power generation, grid, load, and storage is constructed. The mechanism model is calibrated using the multi-scale feature dataset to generate a digital twin model that is synchronously mapped to the physical power system in real time. Multiple planning scenarios are generated in the digital twin model, and a set of uncertainties is associated with each planning scenario; For each planning scenario, multi-objective collaborative planning calculations are performed in the digital twin model to generate a set of candidate planning schemes; The results of big data analysis and mining are used to evaluate the candidate planning scheme set from multiple dimensions, and the optimal collaborative planning scheme is selected.

2. The source-grid-load-storage coordinated planning method for a new type of power system according to claim 1, characterized in that, Big data analysis and mining include: A hierarchical clustering algorithm based on dynamic time warping is used to divide the minute-level load data and the minute-level output data of each new energy power station into time-series patterns, and extract the typical daily load curve, typical daily output curve and probability distribution corresponding to each typical scenario. A feature selection method based on mutual information is adopted, with the new energy output sequence as the target variable and the multi-source meteorological data sequence as the feature variable. The mutual information value between each meteorological factor and the output sequence is calculated, and the dominant meteorological factors are selected by sorting the mutual information values. The transfer entropy method is used to identify the lag correlation between the dominant meteorological factors and the output sequence. The extracted typical scene curves, probability distributions, and identified dominant meteorological factors and their lag correlations are fused and encoded to construct a multi-scale feature dataset of source-grid-load-storage that includes time-domain features, frequency-domain features, and probability distribution features.

3. The source-grid-load-storage coordinated planning method for a new type of power system according to claim 1, characterized in that, The mechanistic model includes four interconnected and coupled sub-mechanistic models, specifically: The new energy mechanism sub-model is used to describe the active-reactive coupling output characteristics of new energy units. It embeds the irradiance-temperature-power conversion function and the maximum power point tracking control logic. It is used to drive the meteorological input data of new energy power plants and output the time-series predicted values ​​of active power and reactive power. The power flow mechanism sub-model is used to describe the power transmission and voltage distribution characteristics of the power transmission and distribution network. It adopts the node power balance equation in polar coordinate form and introduces the adjustment logic of the on-load tap changer of the transformer and the automatic switching logic of the parallel capacitor bank. It takes the injected power of each node as input and outputs the voltage amplitude and phase angle of each node, and the active power and reactive power of each branch. The load response mechanism sub-model is used to describe the demand response characteristics of adjustable loads. It includes a load elasticity coefficient matrix based on real-time electricity prices and load interruption and load transfer contract logic based on demand-side response contracts. It takes electricity price signals and control instructions as inputs and outputs the actual response power of load nodes. The energy storage mechanism sub-model is used to describe the charging and discharging dynamic characteristics of energy storage devices, including the charging and discharging efficiency coefficient, self-discharge rate and cycle life decay factor. It takes the charging and discharging power command as input and outputs the recursive value of the state of charge time series and the available capacity. The four mechanism sub-models are linked by coupling variables, which include at least: the voltage amplitude and phase angle of each node, the power of each branch, and the injected power of each node, forming a hybrid driving mechanism model in the form of a hybrid integer nonlinear programming.

4. The source-grid-load-storage coordinated planning method for a new type of power system according to claim 1, characterized in that, Layered parameter calibration includes: First layer: Using historical power output data from renewable energy power plants, identify the power conversion coefficient and time delay parameters in the renewable energy power output mechanism sub-model online; The second layer: using historical voltage and power data of each node in the distribution network, online estimation of line impedance parameters and transformer turns ratio parameters in the power flow mechanism sub-model; The third layer: After locking the identification results of the first and second layers, the charging and discharging efficiency and capacity decay coefficient in the energy storage mechanism sub-model are identified online using the historical operating data of the energy storage system. The parameters after hierarchical calibration are substituted into the mechanistic model to generate the digital twin model, which has a model error feedback interface for rolling correction.

5. The source-grid-load-storage coordinated planning method for a new type of power system according to claim 1, characterized in that, Multiple planning scenarios are generated in the digital twin model, and a set of uncertainties is associated with each planning scenario, specifically including: Simulate a scenario of increased new energy penetration by modifying installed capacity parameters; By modifying the load baseline curve and overlaying the load morphology variation pattern extracted from a multi-scale feature dataset, a load growth scenario is simulated. Different energy storage configuration schemes can be simulated by modifying the capacity and power parameters of the energy storage device at a specified node. Based on the uncertainty intervals obtained from big data analysis and mining, each planning scenario is associated with a set of uncertainties represented in the form of interval numbers or fuzzy numbers.

6. The source-grid-load-storage coordinated planning method for a new type of power system according to claim 1, characterized in that, Multi-objective collaborative planning calculations include: The system aims to minimize the total life-cycle cost of the system and maximize the annual consumption of renewable energy as the synergistic optimization objectives, and uses nodal power balance constraints, line power flow limit constraints, new energy output uncertainty constraints, energy storage state of charge constraints, and adjustable load response capacity constraints as operational constraints. The above planning model is solved by a multi-objective evolutionary algorithm, resulting in a set of candidate planning schemes. Any scheme in this set of candidate planning schemes cannot further optimize any objective without degrading the other objective. During the solution process, the model error feedback interface of the digital twin model is used to receive measured data for rolling parameter correction.

7. The source-grid-load-storage coordinated planning method for a new type of power system according to claim 1, characterized in that, Multi-dimensional assessment includes: The total lifecycle cost of each candidate planning scheme is used as the first dimension of evaluation value; In the digital twin model corresponding to each of the candidate planning schemes, N-1 fault scans are performed one by one, and the resulting power shortage probability and expected power shortage amount are used as the second dimension evaluation value. For each candidate planning scheme associated with an uncertainty set, the extreme value difference between the two optimization objectives under each candidate planning scheme within the range of the uncertainty set is output, and the extreme value difference is used as the third dimension evaluation value. The normalized values ​​of each candidate planning scheme in the first, second, and third dimensions are used as inputs to the fuzzy hierarchical analysis method, and the subjective weights of each dimension are output. The normalized values ​​are used as input to the entropy weight method, and the objective weights of each dimension are output. The subjective weights and objective weights are multiplied and combined to output the comprehensive weight; Based on the comprehensive weight, each candidate planning scheme is weighted and ranked, and the candidate planning scheme with the highest ranking is output as the optimal collaborative planning scheme.

8. A source-grid-load-storage coordinated planning system for a new type of power system, configured to execute a source-grid-load-storage coordinated planning method for a new type of power system as described in any one of claims 1-7, characterized in that, include: The multi-scale feature extraction module is configured to be connected to the data acquisition and monitoring control system of the new power system, acquire multi-year continuous operation data, run big data analysis and mining process, and output multi-scale feature datasets of source, grid, load and storage. The mechanism model construction module is configured to call a pre-set power system component parameter library to establish mathematical equations for new energy generating units, power transmission and distribution networks, adjustable loads and energy storage devices, and to form a mechanism model in the form of mixed integer nonlinear programming by coupling variables. The twin mapping module is configured to receive the multi-scale feature dataset to perform hierarchical parameter identification on the mechanism model, establish an error feedback channel between the identification results and the real-time measurement values ​​of the physical system, and output a dynamically corrected digital twin model. The scenario construction module is configured to inject adjustments to the penetration rate of new energy sources, the load growth rate, and modifications to the energy storage configuration into the digital twin model in the form of parameter perturbation, generate multiple planning scenarios, and add interval constraints to each planning scenario based on the uncertainty boundary obtained from big data analysis and mining. The collaborative optimization solution module is configured to perform a multi-objective evolutionary search under each planning scenario, with a specified bi-objective function as the optimization direction and the operational constraint boundary provided by the digital twin model as the feasible region, to generate a set of candidate planning schemes. The solution decision module is configured to receive the set of candidate planning solutions, call historical statistical features obtained from big data analysis and mining to construct a multi-dimensional evaluation matrix, and output the optimal collaborative planning solution after comprehensive evaluation and screening.

Images