Differentiated medium and long term contract deviation evaluation method based on load prediction precision
By constructing a differentiated assessment mechanism based on load forecasting accuracy, the unfairness caused by differences in user forecasting capabilities in medium- and long-term power transactions has been resolved. This has enabled dynamic evaluation and positive incentives for load forecasting capabilities, thereby improving the operational efficiency and fairness of the power market.
Patent Information
- Application Number
- CN202510989778.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-17
- Publication Date
- 2025-11-11
AI Technical Summary
The existing medium- and long-term power trading deviation assessment mechanism fails to fully consider the differences in forecasting capabilities among users, resulting in some users bearing unfair responsibilities beyond their control. Furthermore, it lacks positive incentives for high-precision behavior, making it difficult to effectively drive users to optimize their forecasting capabilities.
By quantifying users' historical prediction behavior characteristics, a tiered assessment rule is constructed that is dynamically linked to prediction capabilities. The prediction accuracy index is calculated using a rolling weighted average method, and allowable deviation thresholds and penalty coefficients are dynamically set. The assessment results are adjusted in conjunction with a real-time feedback mechanism to achieve differentiated deviation responsibility identification.
It enables an objective assessment of users' load forecasting capabilities, avoids excessive penalties for users with low accuracy, improves the fairness of the assessment process, and enhances the overall system load forecasting level through positive incentives, driving users to continuously optimize forecasting models.
Smart Images
Figure CN120930859A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power market operation technology, and in particular to a method for evaluating the deviation of differentiated medium- and long-term contracts based on the accuracy of load forecasting. Background Technology
[0002] With the continuous advancement of the electricity market mechanism, medium- and long-term electricity trading, as an important link connecting planned dispatch and the spot market, has become a key component of the electricity trading system. Under this model, power generators and power consumers agree on the electricity delivery plan through contracts, aiming to achieve coordinated operation of market-based trading and system balance. Due to the certain fluctuations in load and the influence of various factors on the forecasting process, deviations inevitably occur in the actual execution of electricity. Therefore, it is necessary to establish a scientific and reasonable deviation assessment mechanism to ensure the stable operation of the system.
[0003] Currently, common deviation assessment mechanisms mainly include fixed threshold penalties, market-based deviation settlement, and planned electricity coverage. These methods have significant shortcomings in practical applications: on the one hand, fixed threshold methods fail to fully consider the differences in forecasting capabilities among users, which can easily lead to some users bearing unfair responsibilities beyond their control; on the other hand, although market-based methods introduce a price mechanism, they do not utilize users' historical forecasting performance and lack positive incentives for high-precision behavior. At the same time, existing mechanisms generally suffer from insufficient incentive guidance and lack of dynamic adjustment, making it difficult to effectively drive users to continuously optimize their forecasting capabilities. Summary of the Invention
[0004] This invention provides a differentiated medium- and long-term contract deviation assessment method based on load forecasting accuracy. By quantifying users' historical forecasting behavior characteristics, it constructs a hierarchical assessment rule that is dynamically linked to forecasting capabilities, and adjusts the assessment results in conjunction with a real-time feedback mechanism. This enables differentiated deviation responsibility identification for entities with different forecasting capabilities, thereby improving the overall system operating efficiency and market participation fairness.
[0005] A method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy includes the following steps:
[0006] S1: Obtain the user's contracted electricity volume, actual electricity consumption and predicted electricity volume within the historical window period. Use the rolling weighted average method to calculate the user's monthly prediction accuracy index. Use the time decay weight to reflect the recent prediction performance. Finally, classify the user's prediction accuracy level according to the accuracy index.
[0007] S2, dynamically sets the allowable deviation threshold based on the user's prediction accuracy level. The allowable deviation threshold is inversely linked to the accuracy level, and sets the penalty coefficient for positive and negative deviations. The penalty intensity is divided according to the level, and a graded assessment rule matching the user's prediction ability is constructed.
[0008] S3 calculates the user's monthly positive and negative deviation electricity consumption and settles accounts based on the penalty coefficient and deviation benchmark price corresponding to the level. At the same time, it sets up an accuracy tracking feedback mechanism to update the user level monthly and dynamically adjust the penalty parameters as the accuracy continues to improve, so as to realize dynamic guidance and positive incentives for deviation responsibility.
[0009] Optionally, S1 includes:
[0010] S11, Based on the power trading platform database or the user electricity consumption information collection system, obtain the contracted electricity volume Q of user u within the historical window period T (default 12 months). C,u,t Actual electricity consumption Q A,u,t Predicted power Q F,u,t ;
[0011] S12, calculate the accuracy index AI of user u in month k using the rolling weighted average method. u (k);
[0012] S13, the accuracy index AI u (k) is mapped to prediction accuracy levels, including Level A (Excellent), Level B (Good), Level C (Good), Level D (Acceptable), and Level E (Needs Improvement).
[0013] Optionally, the accuracy index AI u (k) is represented as:
[0014]
[0015] Among them, w t This represents the time decay weight.
[0016] Optionally, the time decay weight w t The attenuation factor is set and expressed as follows:
[0017] w t =λ k-t ;
[0018] Where λ is the attenuation factor.
[0019] Optionally, S13 includes:
[0020] If AI u If (k)∈[0.95,1], then the prediction accuracy level is Level A (Excellent);
[0021] If AI u If (k)∈[0.85,0.95), then the prediction accuracy level is level B (excellent);
[0022] If AI uIf (k)∈[0.75,0.85), then the prediction accuracy level is level C (good);
[0023] If AI u If (k)∈[0.60,0.75), then the prediction accuracy level is level D (qualified);
[0024] If AI u If (k)∈[0,0.60), then the prediction accuracy level is level E (to be improved).
[0025] Optionally, S2 includes:
[0026] S21, Allowable deviation threshold θ u It is inversely linked to the prediction accuracy level, and a deviation threshold θ is calculated. u ;
[0027] S22, Set the positive and negative deviation penalty coefficient K according to the user's prediction accuracy level. p K n Specifically, it includes:
[0028] For level A, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 0.8. n It is 0.5;
[0029] For grade B, a positive deviation penalty coefficient K is set. p The negative deviation penalty coefficient K is 1.0. n It is 0.8;
[0030] For grade C, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 1.2. n It is 1.0;
[0031] For level D, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 1.5. n It is 1.2;
[0032] For level E, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 2.0. n It is 1.5.
[0033] Optionally, the deviation threshold θ u Represented as:
[0034] θ u =θ base ×(1+β·Rank u );
[0035] Where, θ base As the baseline threshold, Ranku This is the grade coefficient.
[0036] Optionally, S3 includes:
[0037] S31, calculate the deviation in electricity consumption for user u in month k, expressed as:
[0038] ΔQ u =Q A,u -Q C,u ;
[0039] S32, when ΔQ u >θ u Q C,u When the positive deviation penalty cost is expressed as:
[0040] C p,u =K p ·P penalty ·(ΔQ u -θ u Q C,u );
[0041] Among them, P penalty As a penalty benchmark price;
[0042] S33, when ΔQ u <-θ u Q C,u When the negative deviation penalty cost is expressed as:
[0043] C n,u =K n ·P penalty ·(-θ u Q C,u -ΔQ u );
[0044] S34, sets up a precision tracking feedback mechanism, and updates the AI monthly. u (k) and reclassify the levels;
[0045] S35, if the user's prediction accuracy level improves for three consecutive months (e.g., D→C), then K p K n Temporarily lowered by 0.1.
[0046] The beneficial effects of this invention are:
[0047] This invention introduces a rolling weighted average accuracy index calculation method and combines it with a time decay weight mechanism to quantitatively analyze users' historical prediction behavior. This enables an objective assessment of users' load prediction capabilities and allows users to be divided into different prediction accuracy levels. This effectively avoids the excessive punishment of low-accuracy subjects by the traditional one-size-fits-all assessment mechanism and improves the fairness and relevance of the assessment process.
[0048] This invention constructs a dynamic mapping relationship between prediction level, deviation threshold, and penalty coefficient, and combines it with an accuracy tracking feedback mechanism to incentivize users who improve their prediction capabilities by lowering the penalty coefficient. This guides users to continuously optimize their prediction models, improves the overall load prediction level of the entire system, and ultimately achieves differentiated management and positive incentives for determining responsibility for deviations in long-term transactions in the power market. Attached Figure Description
[0049] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0050] Figure 1 This is a schematic diagram of the evaluation method flow according to an embodiment of the present invention. Detailed Implementation
[0051] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.
[0052] like Figure 1 As shown, a differentiated medium- and long-term contract deviation assessment method based on load forecasting accuracy includes the following steps:
[0053] Step 1: Quantitative Modeling of Load Forecasting Accuracy
[0054] Based on the power trading platform database or the user electricity consumption information collection system, enter the key data: the contracted electricity volume Q of user u within the historical window period T (default 12 months). C,u,t Actual electricity consumption Q A,u,t Predicted power Q F,u,t .
[0055] The accuracy index AI of user u in month k is calculated using the rolling weighted average method. u (k):
[0056]
[0057] w t =λ k-t (λ=0.9) (2-2)
[0058] w tλ is the time decay weight, and λ is the decay factor, which controls the decay rate of historical data weight. The smaller λ is, the higher the weight of recent data, thus enhancing the timeliness of trend prediction.
[0059] AI u (k)∈[0,1], a larger value indicates higher precision, according to AI u The (k) value categorizes users into five AE levels, as shown in Table 1.
[0060] Table 1 Prediction Accuracy Levels
[0061] grade A (Excellence) B (Excellent) C (Good) D (Pass) E (Needs improvement) AI scope [0.95,1] [0,85,0.95] [0.75-0.85] [0.60-0.75] [0,0.60]
[0062] Step 2: Dynamically generate differentiated assessment rules:
[0063] Dynamic deviation threshold setting:
[0064] Allowable deviation threshold θ u It is inversely linked to the accuracy level:
[0065] θ u =θ base ×(1+β·Rank u (2-3)
[0066] θ base The baseline threshold is ±4%, referencing Guangdong rules.
[0067] Rank u The grade coefficients are shown in Table 2.
[0068] Table 2 Grade Coefficients
[0069] A B C D E 0.5 0.25 0 -0.25 -0.5
[0070] The penalty coefficient hierarchical mapping is shown in Table 3;
[0071] Table 3 Penalty Coefficient Classification
[0072] grade A B C D E <![CDATA[Positive deviation penalty coefficient K p > 0.8 1.0 1.2 1.5 2.0 <![CDATA[Negative deviation penalty coefficient K n > 0.5 0.8 1.0 1.2 1.5
[0073] Step 3: Deviation Settlement and Dynamic Feedback
[0074] User u's deviation in electricity consumption in month k:
[0075] ΔQ u =Q A,u -Q C,u (2-4)
[0076] Positive deviation (ΔQ) u >θ u Q C,u ):
[0077] Cp,u =K p ·P penalty ·(ΔQ u -θ u Q C,u (2-5)
[0078] Negative deviation (ΔQ) u <-θ u Q C,u ):
[0079] C n,u =K n ·P penalty ·(-θ u Q C,u -ΔQ u (2-6)
[0080] Among them, P penalty To penalize the benchmark price (such as twice the monthly bidding price difference);
[0081] Set up an accuracy tracking feedback mechanism. Update the AI monthly. u (k) And reclassify the levels, driving users to actively optimize the prediction model. If the user's accuracy improves for three consecutive months (e.g., from D to C), K p K n The rate was temporarily lowered by 0.1 to strengthen behavioral guidance.
[0082] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.
[0083] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for assessing deviations in medium- and long-term contracts based on load forecasting accuracy, characterized in that, Includes the following steps: S1: Obtain the user's contracted electricity volume, actual electricity consumption and predicted electricity volume within the historical window period. Use the rolling weighted average method to calculate the user's monthly prediction accuracy index. Use the time decay weight to reflect the recent prediction performance. Finally, classify the user's prediction accuracy level according to the accuracy index. S2, dynamically sets the allowable deviation threshold based on the user's prediction accuracy level. The allowable deviation threshold is inversely linked to the accuracy level, and sets the penalty coefficient for positive and negative deviations. The penalty intensity is divided according to the level, and a graded assessment rule matching the user's prediction ability is constructed. S3 calculates the user's monthly positive and negative deviation electricity consumption and settles accounts based on the penalty coefficient and deviation benchmark price corresponding to the level. At the same time, it sets up an accuracy tracking feedback mechanism to update the user level monthly and dynamically adjust the penalty parameters as the accuracy continues to improve, so as to realize dynamic guidance and positive incentives for deviation responsibility.
2. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 1, characterized in that, S1 includes: S11, Based on the power trading platform database or the user electricity consumption information collection system, obtain the contracted electricity volume Q of user u within the historical window period T. C,u,t Actual electricity consumption Q A,u,t Predicted power Q F,u,t ; S12, calculate the accuracy index AI of user u in month k using the rolling weighted average method. u (k); S13, the accuracy index AI u (k) is mapped to prediction accuracy levels, including level A, level B, level C, level D, and level E.
3. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 2, characterized in that, The accuracy index AI u (k) is represented as: Among them, w t This represents the time decay weight.
4. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 3, characterized in that, The time decay weight w t The attenuation factor is set and expressed as follows: w t =λ k-t ; Where λ is the attenuation factor.
5. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 4, characterized in that, S13 includes: If AI u If (k)∈[0.95,1], then the prediction accuracy level is level A; If AI u If (k)∈[0.85,0.95), then the prediction accuracy level is level B; If AI u If (k)∈[0.75,0.85), then the prediction accuracy level is level C; If AI u If (k)∈[0.60,0.75), then the prediction accuracy level is level D; If AI u If (k)∈[0,0.60), then the prediction accuracy level is level E.
6. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 5, characterized in that, S2 includes: S21, Allowable deviation threshold θ u It is inversely linked to the prediction accuracy level, and a deviation threshold θ is calculated. u ; S22, Set the positive and negative deviation penalty coefficient K according to the user's prediction accuracy level. p K n Specifically, it includes: For level A, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 0.
8. n It is 0.5; For grade B, a positive deviation penalty coefficient K is set. p The negative deviation penalty coefficient K is 1.
0. n It is 0.8; For grade C, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 1.
2. n It is 1.0; For level D, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 1.
5. n It is 1.2; For level E, set a positive deviation penalty coefficient K. p The negative deviation penalty coefficient K is 2.
0. n It is 1.
5.
7. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 6, characterized in that, The deviation threshold θ u Represented as: i u =θ base ×(1+β·Rank u ); Where, θ base As the baseline threshold, Rank u This is the grade coefficient.
8. The method for assessing the deviation of differentiated medium- and long-term contracts based on load forecasting accuracy according to claim 7, characterized in that, S3 includes: S31, calculate the deviation in electricity consumption for user u in month k, expressed as: ΔQ u =Q A,u -Q C,u ; S32, when ΔQ u >θ u Q C,u When the positive deviation penalty cost is expressed as: C p,u =K p ·P penalty ·(ΔQ u -θ u Q C,u ); Among them, P penalty As a penalty benchmark price; S33, when ΔQ u <-θ u Q C,u When the negative deviation penalty cost is expressed as: C n,u =K n ·P penalty ·(-θ u Q C,u -ΔQ u ); S34, sets up a precision tracking feedback mechanism, and updates the AI monthly. u (k) and reclassify the levels; S35, if the user's prediction accuracy level improves for three consecutive months, then K p K n Temporarily lowered by 0.1.