A cross-provincial power purchase matching scheduling method based on supply guarantee index

Abstract

The application discloses a cross-provincial power purchase matching scheduling method based on a supply guarantee index, and belongs to the technical field of power market scheduling. The application first collects four-dimensional eight-class supply guarantee indexes and carries out standardized preprocessing; adopts AHP+entropy weight combination weighting to calculate dynamic weights, generates a comprehensive supply guarantee index, and divides three-level scheduling intervals; maps the index and cross-provincial power purchase scheduling parameters through a pre-trained BP neural network; adopts a TOPSIS-entropy weight method to quantize power purchase priorities when multiple provinces are simultaneously short of power; and evaluates scheduling efficiency based on three-dimensional indexes after scheduling, and iteratively optimizes index weights and a neural network model in a closed loop in a low-score scene. The application constructs a dynamic quantitative four-dimensional supply guarantee index system, automatically matches power purchase demand through a neural network, quantizes multi-provincial power purchase distribution priorities, and realizes adaptive optimization of a scheduling model.

Description

Technical Field

[0001] This invention belongs to the field of inter-provincial power market transaction and dispatch technology, specifically involving an inter-provincial power purchase matching and dispatch method based on a multi-dimensional quantitative supply guarantee index. It is applicable to the inter-regional power supply guarantee and intelligent dispatch scenario for emergency power purchase in the high-penetration area of ​​the "West-to-East Power Transmission" renewable energy in the five northwestern provinces. Background Technology

[0002] The Northwest region is a core development base for new energy in my country and a key destination for the West-to-East Power Transmission Project. The energy endowments and load scales of the provinces within the region vary significantly: Gansu and Qinghai have large installed capacity of new energy, while Shaanxi and Ningxia have concentrated industrial loads. Situations such as winter cold waves and summer solar PV surges highlight the imbalance between inter-provincial power supply and demand, necessitating inter-provincial power purchases to achieve regional energy mutual assistance and ensure stable electricity supply for both residential and industrial use. Inter-provincial power supply and purchase scheduling is a core technical means for ensuring energy security and the consumption of new energy, and it represents a key research direction for optimizing regional power market scheduling. Currently, inter-provincial power supply dispatching in Northwest China mainly adopts an administrative qualitative assessment + fixed quota allocation model. Existing technologies only use a single power supply adequacy index to roughly assess the power supply status of each province, relying on the manual experience of dispatchers to determine the power shortage level and manually setting inter-provincial power purchase quotas and transaction periods. Some studies have introduced simple two-dimensional power supply indicators, but the indicator weights are fixed and cannot adapt to dynamic conditions such as extreme cold waves and drastic fluctuations in wind and solar power output. When multiple provinces experience simultaneous power shortages and inter-regional transmission channels are congested, power purchase resources are manually allocated based solely on the vague principle of "people's livelihood first," without quantitative priority ranking standards. After dispatching, only the power supply effect is qualitatively described, lacking quantitative performance evaluation methods and unable to optimize dispatching strategies in reverse. The shortcomings of existing technology: 1. The assessment of power supply status is based on a single, qualitative dimension and lacks a dynamic quantitative index system. The existing scheme only focuses on power supply reliability and does not include dimensions such as new energy low-carbon, economic cost of electricity purchase, and social and livelihood losses. Moreover, the weight of the indicators is fixed, and the assessment is distorted under the conditions of fluctuating wind and solar power output and extreme cold waves. It cannot accurately reflect the real supply pressure in the region, resulting in large deviations in the calculation of cross-provincial electricity purchase demand. 2. There is no standardized mapping relationship between supply guarantee demand and inter-provincial electricity purchase parameters. The purchase volume, price, and transaction time are matched manually based on experience, resulting in a serious mismatch between supply and demand. Provinces with high supply guarantee pressure have insufficient electricity purchase quotas, while provinces with low pressure have redundant electricity purchases, leading to a waste of channel resources. 3. When multiple provinces experience simultaneous power shortages, there is no quantitative ranking standard for the allocation of purchased electricity resources. With limited transmission capacity, relying on subjective human judgment to allocate electricity can easily lead to problems such as industrial loads crowding out residential electricity use and insufficient priority given to renewable energy, resulting in poor fairness in supply guarantees and inadequate capacity to protect people's livelihoods. 4. Lack of a closed-loop iterative optimization mechanism for scheduling efficiency. The scheduling effect is only reviewed manually after the fact, which cannot quantify response time, electricity purchase cost, and supply-demand matching deviation. It cannot automatically correct the supply guarantee assessment model and electricity purchase matching rules, resulting in slow supply guarantee response speed. Currently, the response time for cross-provincial emergency electricity purchase is generally 4 to 6 hours, and the electricity purchase cost is relatively high. Summary of the Invention

[0003] To address the shortcomings of existing inter-provincial power supply guarantee scheduling in Northwest China, such as qualitative assessment, supply-demand mismatch, lack of quantitative standards for resource allocation, and lack of closed-loop optimization mechanisms, this paper proposes an inter-provincial power purchase matching and scheduling method based on a power supply guarantee index. A four-dimensional dynamic weighted comprehensive power supply guarantee index quantification system is constructed. A nonlinear mapping between the index and power purchase parameters is established through a neural network to quantify the power purchase priorities of multiple provinces. Based on scheduling efficiency, the model achieves adaptive iteration, enabling accurate, fair, and rapid intelligent scheduling of inter-provincial power purchases. The objective of this invention can be achieved through the following technical solutions: A cross-provincial power purchase matching and dispatching method based on a supply guarantee index includes the following steps: S1. Collect multi-dimensional raw power supply indicators from the region and perform standardized preprocessing to construct a standardized power supply indicator dataset; S2. Dynamic weights are calculated using a combination of subjective and objective weighting methods. Based on a standardized supply guarantee indicator dataset, a comprehensive supply guarantee index for each province is generated, and multi-level supply guarantee scheduling intervals are defined. S3. Using a pre-trained neural network mapping model, the comprehensive supply guarantee index is mapped and output as the basic parameters for cross-provincial power purchase and dispatch. S4. When multiple provinces have simultaneous electricity purchase demand and transmission channels are limited, a multi-attribute sorting algorithm is used to determine the priority of electricity purchase resource allocation for each province. S5. Perform cross-provincial power purchase matching and dispatch according to priority, and collect data on the entire dispatch process. S6. Quantitatively evaluate the overall efficiency of this scheduling, and update the supply guarantee index weight and neural network mapping model in a closed loop based on the efficiency evaluation results to complete adaptive optimization. As a further aspect of the present invention, step S1 specifically includes: S11 connects with provincial marketing systems, dispatch SCADA systems, power trading centers, and environmental protection platforms to collect eight underlying indicators across four dimensions: power supply reliability, energy sustainability, economic cost, and social impact. These eight indicators are: load deficit rate, annual cumulative power outage duration, proportion of renewable energy supply, carbon emission reduction from replacing thermal power, emergency power purchase price, transmission cost ratio, residential electricity supply guarantee rate, and industrial shutdown losses. Specifically, the load deficit rate = regional power deficit / total regional electricity demand; the renewable energy supply guarantee ratio = renewable energy supply / total regional power supply guarantee; and the residential electricity supply guarantee rate = actual residential power supply / total residential electricity demand. S12. Use the KNN algorithm to fill in missing values ​​for the indicators and use the 3σ Laida criterion to remove anomalous jump data. S13. Use min-max normalization for unified standardization: Standardized score of positive indicators = actual value of indicator / maximum value of indicator, standardized score of negative indicators = 1 - actual value of indicator / maximum value of indicator. All indicators are uniformly mapped to the range of 0~100 points to form a standardized four-dimensional supply guarantee indicator dataset. As a further aspect of the present invention, step S2 specifically includes: S21. Subjective weights are calculated using the AHP (Analytic Hierarchy Process): Experts compare each of the four dimensions pairwise to construct a judgment matrix and obtain eigenvectors to get the subjective weights. In extreme power shortage warning scenarios, the subjective weight of the power supply reliability dimension is increased to 30%. S22. Calculate the objective weights using the entropy weight method, based on the information entropy formula. The information entropy of each indicator is calculated. The greater the fluctuation of the indicator and the lower the information entropy, the higher the objective weight. When the monthly fluctuation of the proportion of new energy supply exceeds 15%, the objective weight of the energy sustainability dimension is automatically increased to 25%. S23. Dynamic weight matrix integration: Subjective weight accounts for 60% and objective weight accounts for 40% of the weighted integration, and the entire weight matrix is ​​automatically updated every quarter; S24. The standardized scores of each dimension are weighted and summed with their corresponding dynamic weights to generate a comprehensive supply guarantee index with a value of 0 to 100. S25. The Otsu's maximum inter-class variance method is used to divide the dispatch intervals into three levels: index < 60 is the emergency supply interval, 60 ≤ index ≤ 80 is the regular supply interval, and index > 80 is the optimized supply interval; different intervals are bound to differentiated power purchase and dispatch rules: the power purchase volume of the emergency supply interval shall not be less than 90% of the power shortage, priority shall be given to off-peak transmission, and the electricity price shall be increased by 15%; the power purchase volume of the regular supply interval shall match 70% / 90% of the power shortage, and the benchmark electricity price shall be applied; the power purchase volume of the optimized supply interval shall give priority to the purchase of new energy power, and the electricity price shall be reduced by 5% / 10%. As a further embodiment of the present invention, the neural network mapping model in step S3 is a three-layer BP neural network, and the model construction and training steps include: S31. Construct a historical sample library: Collect nearly 5 years of power supply guarantee cases from five northwestern provinces to form 100,000 standardized samples. The sample input is the sub-index scores of four dimensions, and the sample output is the cross-provincial power supply volume, power purchase period, and price fluctuation coefficient. The training set and validation set are divided in a 7:3 ratio. The sample library learns a fixed nonlinear association rule: when the standardized score of the power supply guarantee rate for people's livelihood is lower than 80, the demand for cross-provincial power purchase increases by 20%. S32. Network structure configuration: The input layer has 4 neurons corresponding to the four-dimensional sub-item index score, the hidden layer has 10 neurons and uses the ReLU activation function, and the output layer has 3 neurons that output the amount of electricity purchased, the time period for purchasing electricity, and the price fluctuation coefficient, respectively; the training parameters are set with a learning rate of 0.01 and 1000 iterations, using the mean squared error (MSE) as the loss function, and training is carried out until the convergence error on the validation set is ≤0.001; S33. Input the current province's four-dimensional sub-index score into the trained BP neural network, and output the corresponding cross-provincial power purchase basic dispatch parameters. As a further aspect of the present invention, step S4 employs the TOPSIS-entropy weight combination algorithm to achieve priority sorting, the specific steps of which are as follows: S41. Select ranking and evaluation indicators: comprehensive supply guarantee index of each province, proportion of people's livelihood load, and regional industrial output loss; use the entropy weight method to allocate the weight of the indicators, with the proportion of people's livelihood load having a fixed weight of 40%. S42. After standardizing all evaluation indicators, construct a multi-province demand evaluation matrix and use the TOPSIS algorithm to calculate the Euclidean closeness between the supply guarantee demand and the ideal supply guarantee demand of each province. S43. According to the proximity degree, the higher the proximity degree, the higher the priority of allocating inter-provincial power purchase channels and power volume; when the proximity degree of Qinghai is 0.89 and the proximity degree of Gansu is 0.72, the supply guarantee resources will be allocated to Qinghai first. As a further embodiment of the present invention, during the execution of step S5, the system connects in real time to the inter-provincial DC channel SCADA system. If the real-time utilization rate of the channel exceeds 90%, the power purchase plan during the congested period is automatically shifted to the idle period when the channel utilization rate is less than 70%. As a further aspect of the present invention, step S6 specifically includes: S61. Set three-dimensional dispatch efficiency evaluation indicators: response speed indicator is the time from the initiation of power purchase demand to the arrival of power; cost control indicator is the comprehensive power purchase cost per unit of guaranteed power supply; supply and demand balance indicator is the supply and demand deviation rate between the actual power purchase and the power shortage. S62. The entropy method is used to weight the three-dimensional indicators to synthesize a total scheduling efficiency score of 0 to 100 points. A score of 80 points or above is judged as efficient scheduling. S63. If a province's total dispatch efficiency score is below 60 points for three consecutive times, a closed-loop optimization is performed: the weight of the response speed dimension in the comprehensive supply guarantee index is automatically increased to 30%, and the BP neural network mapping model is retrained using the latest dispatch operation data to update the threshold of the power purchase matching dispatch parameters; after closed-loop iterative optimization, the response time for cross-provincial power purchase guarantee is shortened from the original 4-6 hours to less than 2 hours. As a further embodiment of the present invention, the method is deployed on a power trading auxiliary decision-making platform, which integrates three major functional modules: a comprehensive supply guarantee index quantification module, an index-electricity purchase demand mapping module, and an electricity purchase priority and efficiency evaluation module. The beneficial effects of this invention are: This invention constructs a four-dimensional, dynamically weighted comprehensive supply guarantee index to address the shortcomings of existing technologies, such as single-dimensional assessment and distorted static weights. It simultaneously covers eight quantitative indicators across four dimensions: power supply reliability, energy sustainability, economic cost, and social impact. Employing an Alternative Power Hierarchy (AHP) combined with entropy weighting, it balances subjective expert policy needs with objective data fluctuations. In extreme power shortage scenarios, reliability weights are automatically increased; when renewable energy output fluctuates, sustainability weights are automatically adjusted upwards. All indicators are standardized to a score of 0-100, outputting a comprehensive supply guarantee index that can be compared horizontally. Compared to single-indicator assessments, the accuracy of inter-provincial supply guarantee status comparisons is improved, precisely quantifying supply gaps in each province and reducing deviations in electricity demand calculations from the source. A backpropagation (BP) neural network is used to automatically map the supply guarantee index to inter-provincial power purchase parameters, solving the problem of supply-demand mismatch caused by manual experience-based matching. A three-layer BP neural network is trained based on 100,000 real supply guarantee samples from five northwestern provinces over five years. It automatically fits the non-linear relationship between the index score and the purchased electricity volume, transaction period, and price fluctuation coefficient. Combined with the Otsu algorithm, three levels of standardized power purchase scheduling rules are defined. No manual calculation is required, improving the accuracy of the model's recommended power purchase parameters, enhancing the matching degree between the supply guarantee index and power purchase demand, avoiding redundant power purchases or insufficient quotas, fully utilizing inter-regional transmission channels, and reducing resource waste. This invention employs the TOPSIS-Entropy Weight Combination Algorithm to quantify electricity purchase priorities, addressing the shortcomings of unfair manual resource allocation and insufficient public welfare guarantees during power shortages in multiple provinces. It sets a maximum weight of 40% for public welfare load and constructs a ranking index based on a comprehensive supply guarantee index and industrial losses. By closely quantifying the urgency of supply guarantee in each province, it automatically generates an electricity resource allocation sequence. Compared to subjective manual allocation, it prioritizes areas with concentrated public welfare loads, resulting in a higher stability rate of public welfare electricity guarantee in extreme scenarios, thus balancing supply fairness with social stability. A closed-loop optimization mechanism of "scheduling-evaluation-iteration" is established to address the shortcomings of existing technologies, such as the inability to automatically optimize the scheduling model, slow response speed, and high electricity purchase cost. Three-dimensional quantitative performance indicators are set for response speed, unit electricity purchase cost, and supply-demand deviation, and the total scheduling score is synthesized through the entropy method. In scenarios with continuous low scores, the weight of the supply guarantee index is automatically adjusted and the neural network mapping model is retrained. Attached Figure Description The invention will now be further described with reference to the accompanying drawings. Figure 1 This is a schematic diagram of the principle of a cross-provincial power purchase matching and scheduling method based on a supply guarantee index in this invention. Detailed Implementation The present invention will be further described in detail below with reference to specific embodiments. These embodiments are intended to fully disclose the technical solutions of the present invention, enabling those skilled in the art to implement the present invention, and are not intended to limit the scope of protection of the present invention. Example: Gansu-Qinghai inter-provincial winter cold wave power supply dispatch (typical power shortage scenario in Northwest China in January 2026) 4.1.1 Basic Data Acquisition (Execution Step S1) During the cold wave scenario, eight standardized indicators were simultaneously collected from Gansu and Qinghai provinces. The normalized scores for each indicator are shown in Table 1. Power supply reliability Load deficit rate, power outage duration 62、58 45、42 Energy sustainability The proportion of new energy supply and carbon emission reduction 70、68 82、80 Economic Costs Emergency power purchase price and the proportion of transmission costs 55、60 68、72 Social impact People's livelihood security rate, industrial production stoppage losses 70、65 88、90 Data preprocessing: No missing data or abnormal jumps are detected. Min-max normalization is performed directly to form a standardized dataset. Calculation of the comprehensive supply guarantee index (Execution step S2) AHP Subjective Weights (Cold Wave Emergency Scenario): Power Supply Reliability 0.3, Sustainability 0.2, Economic Efficiency 0.2, Social Impact 0.3; The objective weights for the entropy weight method are: reliability 0.28, sustainability 0.25, economy 0.22, and social impact 0.25. Incorporate dynamic weighting (subjective 60% + objective 40%): Reliability: 0.3×0.6+0.28×0.4=0.292; Sustainability: 0.2×0.6+0.25×0.4=0.22; Economic efficiency: 0.2×0.6+0.22×0.4=0.208; Social impact: 0.3×0.6+0.25×0.4=0.28; Exponential weighted calculation: Gansu's comprehensive index = (60 × 0.292) + (69 × 0.22) + (57.5 × 0.208) + (67.5 × 0.28) = 63.4 points; Qinghai's comprehensive index = (43.5 × 0.292) + (81 × 0.22) + (70 × 0.208) + (89 × 0.28) = 68.7 points; Range determination: The indices of both provinces are in the normal supply guarantee range of 60-80 points. The BP neural network outputs the power purchase and dispatch parameters (step S3). Input the average scores of the four dimensions of the two provinces into the pre-trained BP model, and output the scheduling parameters: Gansu: 78% shortfall in inter-provincial power purchases, mainly relying on flat-rate electricity, with prices rising by 3%; Qinghai: 85% shortfall in inter-provincial electricity purchases, mainly during off-peak hours, electricity prices increased by 5%; In accordance with standard supply guarantee rules: the purchased electricity volume will be matched with a gap of 70% to 90%, and the market electricity price will be slightly increased. Priority sorting (step S4) Ranking indicators: People's livelihood load (weight 40%), comprehensive supply guarantee index (30%), industrial losses (30%). The load on residential electricity consumption accounts for 65% in Qinghai and 52% in Gansu. According to TOPSIS calculations, the proximity of Qinghai to residential electricity consumption is 0.89, and that of Gansu is 0.72. The dispatch priority is Qinghai > Gansu, and the Qishao DC transmission line capacity will be allocated first. Scheduling execution and performance closed-loop iteration (steps S5 and S6) Dispatch execution: Priority was given to allocating 1.2 million kilowatts of off-peak inter-provincial power purchase quota to Qinghai, and 850,000 kilowatts of regular quota to Gansu. The channel utilization rate remained stable at 82%, with no congestion. Performance metrics collected: The dispatch response time was 1.5 hours, the unit electricity purchase cost was 0.21 yuan / MWh, and the supply-demand deviation rate was 2.1%. Overall performance score calculation: The entropy weighted composite score is 86 points, which is judged as efficient scheduling, and no iterative optimization of the model is required. Results of this dispatch: The electricity supply guarantee rate for residential use in both provinces reached 98.5%, there were no large-scale industrial power outages, surplus hydropower from Qinghai was absorbed, and the amount of water and solar power wasted in the region was reduced by 12 million kilowatt-hours. Comparative Example: Traditional manual qualitative scheduling scheme (same as cold wave scenario) Using existing technology for manual scheduling relies solely on the load deficit rate as a single indicator to assess supply pressure, lacking quantitative indices and priority algorithms. Only two provinces were identified as having power shortages, and the power purchase quotas were roughly allocated manually: 1.3 million kilowatts for Gansu and 750,000 kilowatts for Qinghai; The power grid failed to differentiate between the weight of the load on the people's livelihood and prioritized the allocation of electricity to Gansu, where the industrial load is concentrated, while the power shortage in Qinghai was not fully covered. The dispatch response time is 5 hours. During periods of channel congestion, a large amount of electricity is purchased, and the channel utilization rate is 96%, resulting in power abandonment due to congestion. The unit electricity purchase cost is 0.25 yuan / MWh, the supply and demand deviation rate is 13.2%, the electricity guarantee rate for people's livelihood in Qinghai is only 91%, small-scale power rationing for residents has occurred, and the hourly loss for industrial enterprises is about 21 million yuan. Comparison of Example Data and Comparative Data Supply demand matching degree 87% 56% 31% increase Four-dimensional exponential analysis accurately quantifies the gap, and neural network nonlinear fitting of power purchase parameters eliminates the bias of manual estimation. Scheduling response time 1.5h 5h Shortened by 3.5 hours A standardized, end-to-end algorithm automatically calculates indices, electricity purchase parameters, and priorities, eliminating the need for manual analysis, calculation, and consultation. Unit inter-provincial electricity purchase cost 0.21 yuan / MWh 0.25 yuan / MWh 16% lower The algorithm automatically matches low-price, off-peak periods and low-congestion corridors, dynamically optimizing electricity purchase times and transmission routes. Electricity supply guarantee rate for residential use 98.5% 91% Increased by 7.5% The TOPSIS algorithm sets the weight of residential load to the highest level, and prioritizes the allocation of electricity purchase quotas to residential areas. Supply-demand deviation rate 2.1% 13.2% Reduced by 11.1% The dynamic supply guarantee index adjusts the gap assessment in real time, and the neural network outputs precise electricity purchase quotas to avoid redundancy / insufficiency. Multi-scenario verification supplement Extreme early warning scenario (comprehensive index 52 points, <60 points): Implement emergency supply guarantee rules, purchase electricity to cover 92% of the shortage, increase electricity price by 15%, and dispatch response time is 1.8 hours; Optimize supply guarantee scenarios (comprehensive index 86 points, >80 points): prioritize the procurement of photovoltaic and wind power, reduce electricity prices by 8%, and increase the proportion of regional renewable energy supply to 26%; Continuous inefficient scheduling closed-loop iteration verification: A province scored 55 points in efficiency three times in a row. The system automatically increased the response speed weight to 30%. After retraining the BP model, the subsequent scheduling response time was shortened from 3.8h to 1.7h, verifying the effectiveness of the closed-loop optimization mechanism. In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. The above description is merely an example and illustration of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the invention or exceed the scope defined in the claims, they should all fall within the protection scope of the present invention.

Claims

1. A cross-provincial power purchase matching and dispatching method based on a supply guarantee index, characterized in that, Includes the following steps: S1. Collect multi-dimensional raw power supply indicators from the region and perform standardized preprocessing to construct a standardized power supply indicator dataset; S2. Dynamic weights are calculated using a combination of subjective and objective weighting methods. Based on the standardized supply guarantee indicator dataset, a comprehensive supply guarantee index for each province is generated, and multi-level supply guarantee scheduling intervals are divided. S3. Using a pre-trained neural network mapping model, the comprehensive supply guarantee index is mapped and output as the basic parameters for cross-provincial power purchase and dispatch. S4. When multiple provinces have simultaneous electricity purchase demand and transmission channels are limited, a multi-attribute sorting algorithm is used to determine the priority of electricity purchase resource allocation for each province. S5. Perform cross-provincial power purchase matching and dispatch according to priority, and collect data on the entire dispatch process. S6. Quantitatively evaluate the overall efficiency of this scheduling, and update the supply guarantee index weight and neural network mapping model in a closed loop based on the efficiency evaluation results to complete adaptive optimization.

2. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, Step S1 specifically includes: S11 connects with provincial marketing systems, dispatch SCADA systems, power trading centers, and environmental protection platforms to collect eight underlying indicators across four dimensions: power supply reliability, energy sustainability, economic cost, and social impact. These eight indicators are: load deficit rate, annual cumulative power outage duration, proportion of renewable energy supply, carbon emission reduction from replacing thermal power, emergency power purchase price, transmission cost ratio, residential electricity supply guarantee rate, and industrial shutdown losses. Specifically, the load deficit rate = regional power deficit / total regional electricity demand; the renewable energy supply guarantee ratio = renewable energy power supply / total regional power supply guarantee; and the residential electricity supply guarantee rate = actual residential power supply / total residential electricity demand. S12. Use the KNN algorithm to fill in missing values ​​for indicators and use the 3σ Laida criterion to remove anomalous jump data. S13. Use min-max normalization for unified standardization: Standardized score of positive indicators = actual value of indicator / maximum value of indicator, standardized score of negative indicators = 1 - actual value of indicator / maximum value of indicator. All indicators are uniformly mapped to the range of 0~100 points to form a standardized four-dimensional supply guarantee indicator dataset.

3. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, Step S2 specifically includes: S21. Subjective weights are calculated using the AHP (Analytic Hierarchy Process): Experts compare each of the four dimensions pairwise to construct a judgment matrix and obtain eigenvectors to get the subjective weights. In extreme power shortage warning scenarios, the subjective weight of the power supply reliability dimension is increased to 30%. S22. Calculate the objective weights using the entropy weight method, based on the information entropy formula. The information entropy of each indicator is calculated. The greater the fluctuation of the indicator and the lower the information entropy, the higher the objective weight. When the monthly fluctuation of the proportion of new energy supply exceeds 15%, the objective weight of the energy sustainability dimension is automatically increased to 25%. S23. Dynamic weight matrix integration: Subjective weight accounts for 60% and objective weight accounts for 40% of the weighted integration, and the entire weight matrix is ​​automatically updated every quarter; S24. The standardized scores of each dimension are weighted and summed with their corresponding dynamic weights to generate a comprehensive supply guarantee index with a value of 0 to 100. S25. The Otsu's maximum inter-class variance method is used to divide the dispatch intervals into three levels: index < 60 is the emergency supply interval, 60 ≤ index ≤ 80 is the regular supply interval, and index > 80 is the optimized supply interval; different intervals are bound to differentiated power purchase and dispatch rules: the power purchase volume of the emergency supply interval shall not be less than 90% of the power shortage, priority shall be given to off-peak transmission, and the electricity price shall be increased by 15%; the power purchase volume of the regular supply interval shall match 70% / 90% of the power shortage, and the benchmark electricity price shall be implemented; the power purchase volume of the optimized supply interval shall give priority to the purchase of new energy power, and the electricity price shall be reduced by 5% / 10%.

4. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, In step S3, the neural network mapping model is a three-layer BP neural network. The model construction and training steps include: S31. Construct a historical sample library: Collect nearly 5 years of power supply guarantee cases from five northwestern provinces to form 100,000 standardized samples. The sample input is the sub-index scores of four dimensions, and the sample output is the cross-provincial power supply volume, power purchase period, and price fluctuation coefficient. The training set and validation set are divided in a 7:3 ratio. The sample library learns a fixed nonlinear association rule: when the standardized score of the power supply guarantee rate for people's livelihood is lower than 80, the demand for cross-provincial power purchase increases by 20%. S32. Network structure configuration: The input layer has 4 neurons corresponding to the four-dimensional sub-item index score, the hidden layer has 10 neurons and uses the ReLU activation function, and the output layer has 3 neurons that output the amount of electricity purchased, the time period for purchasing electricity, and the price fluctuation coefficient, respectively; the training parameters are set with a learning rate of 0.01 and 1000 iterations, using the mean squared error (MSE) as the loss function, and training is carried out until the convergence error on the validation set is ≤0.001; S33. Input the current province's four-dimensional sub-index score into the trained BP neural network, and output the corresponding cross-provincial power purchase basic dispatch parameters.

5. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, Step S4 uses the TOPSIS-entropy weight combination algorithm to implement priority sorting. The specific steps are as follows: S41. Select ranking and evaluation indicators: comprehensive supply guarantee index of each province, proportion of people's livelihood load, and regional industrial output loss; use the entropy weight method to allocate the weight of the indicators, with the proportion of people's livelihood load having a fixed weight of 40%. S42. After standardizing all evaluation indicators, construct a multi-province demand evaluation matrix and use the TOPSIS algorithm to calculate the Euclidean closeness between the supply guarantee demand and the ideal supply guarantee demand of each province. S43. According to the proximity degree, the higher the proximity degree, the higher the priority of allocating inter-provincial power purchase channels and power volume; when the proximity degree of Qinghai is 0.89 and the proximity degree of Gansu is 0.72, the supply guarantee resources will be allocated to Qinghai first.

6. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, Step S5 involves real-time connection to the inter-provincial DC channel SCADA system during the scheduling process. If the channel's real-time utilization rate exceeds 90%, the power purchase plan for congested periods will be automatically shifted to idle periods when the channel utilization rate is below 70%.

7. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, Step S6 specifically includes: S61. Set three-dimensional dispatch efficiency evaluation indicators: response speed indicator is the time from the initiation of power purchase demand to the arrival of power; cost control indicator is the comprehensive power purchase cost per unit of guaranteed power supply; supply and demand balance indicator is the supply and demand deviation rate between the actual power purchase and the power shortage. S62. The entropy method is used to weight the three-dimensional indicators to synthesize a total scheduling efficiency score of 0 to 100 points. A score of 80 points or above is judged as efficient scheduling. S63. If a province's total dispatch efficiency score is below 60 points for three consecutive times, a closed-loop optimization is performed: the weight of the response speed dimension in the comprehensive supply guarantee index is automatically increased to 30%, and the BP neural network mapping model is retrained using the latest dispatch operation data to update the threshold of the power purchase matching dispatch parameters; after closed-loop iterative optimization, the response time for cross-provincial power purchase guarantee is shortened from the original 4-6 hours to less than 2 hours.

8. The method for cross-provincial power purchase matching and dispatching based on a supply guarantee index according to claim 1, characterized in that, The method is deployed on a power trading auxiliary decision-making platform, which integrates three major functional modules: a comprehensive supply guarantee index quantification module, an index-electricity purchase demand mapping module, and an electricity purchase priority and efficiency evaluation module.

Images