Flood forecasting effect evaluation method based on rolling forecasting mode

Abstract

The invention discloses a flood forecasting effect evaluation method based on a rolling forecasting mode. The flood forecasting effect evaluation method comprises the following steps: step (1), data collection and arrangement: collecting and arranging an actual flood process needing to be evaluated in a certain time period and corresponding rolling forecasting process data; (2) calculating five evaluation indexes including a mean value process certainty coefficient, an average forecast dispersion, an average relative error, a peak value relative error and a forecast coverage rate; step (3), normalizing the indexes; (4) calculating a comprehensive evaluation index: carrying out weighted average on the normalized indexes in the step (3) to obtain the comprehensive evaluation index; the problem that an existing single flood forecasting process evaluation system is not suitable for a rolling forecasting mode in actual production application is solved, and the flood rolling forecasting effect can be comprehensively evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of flood forecasting effect evaluation, and particularly to a flood forecasting effect evaluation method based on a rolling forecasting mode. Background Technique

[0002] Flood forecasting, as an important non-engineering measure for flood control, directly serves flood control and emergency rescue, rational utilization and protection of water resources, construction and management of water conservancy projects, and safe production of industry and agriculture. Conducting flood forecasting effect evaluation can enable application departments to master the forecast error situation, correctly use the forecast results, and also facilitate problem discovery and improvement of forecast service quality. Currently, there is a relatively mature evaluation system for flood forecasting effects. For example, in the "Hydrological Information and Forecasting Specification (GB / T 22482-2008)", it is recommended that the evaluation of flood forecasting accuracy can be described by indicators such as flood peak flow forecast error, flood peak occurrence time error, flood peak lead time, and coefficient of determination. However, these evaluation criteria mainly target a single forecasting process, while in actual production applications, a rolling forecasting mode is often adopted, that is, a flood forecasting process of a certain time length is released at fixed time intervals. For example, the flood process for the next 10 days is released every day. In the rolling forecasting mode, since an actual flood process corresponds to multiple forecast flood processes, the conventional single forecasting process evaluation system can no longer describe the complex rolling forecasting effect. Therefore, there is an urgent need for a flood forecasting evaluation method based on the rolling forecasting mode to comprehensively evaluate the effect of the rolling forecast for a certain flood process and better guide the flood forecasting work in the actual production process. Summary of the Invention

[0003] The purpose of the present invention is to overcome the above deficiencies, provide a flood forecasting effect evaluation method based on a rolling forecasting mode, and solve the problem that the existing single flood forecasting process evaluation system is not applicable to the rolling forecasting mode in actual production applications, so as to comprehensively evaluate the effect of flood rolling forecasting.

[0004] To solve the above technical problems, the technical solution adopted by the present invention is: a flood forecasting effect evaluation method based on a rolling forecasting mode, which includes the following steps:

[0005] Step (1): Data collection and arrangement: Collect and arrange the actual flood process and the corresponding rolling forecast process data within a certain period to be evaluated.

[0006] Step (2): Calculate 5 evaluation indicators: including the coefficient of determination of the mean process, average forecast dispersion, average relative error, peak relative error, and forecast coverage rate.

[0007] Step (3): Index normalization.

[0008] Step (4): Calculate the comprehensive evaluation index: Perform weighted averaging on the normalized indicators in step (3) to obtain the comprehensive evaluation index.

[0009] Preferably, the data in step (1) includes the following:

[0010]

[0011] Among them, Q is the predicted flood process matrix, O is the measured flood process matrix, 1 to m are the flood process time periods, I to n are the number of rolling predictions, and M and N are auxiliary matrices.

[0012] Preferably, in step (2), the mean process certainty coefficient is:

[0013] DC = NSE(Q', O)

[0014]

[0015] Among them, Q’ represents the rolling prediction average process corresponding to the actual flood process, and NSE represents the calculation method of the certainty coefficient, reflecting the degree of coincidence between the prediction process Q' and the measured process O.

[0016] Preferably, in step (2), the average prediction dispersion is:

[0017]

[0018]

[0019] Among them, Max represents taking the maximum value in the series, Min represents taking the minimum value in the series, and DI represents the average dispersion degree of each rolling prediction value in this flood process.

[0020] Preferably, in step (2), the average relative error is:

[0021]

[0022] Among them, RE is the relative error matrix, and MRE is the average relative error of the rolling prediction result.

[0023] Preferably, in step (2), the peak average error is:

[0024]

[0025] Among them, O j is the peak value of the measured flood, j is its time sequence number, REP is the relative error matrix of the flood peak value, and MREP is the peak average error.

[0026] Preferably, in the step (2), the forecast coverage rate is:

[0027]

[0028]

[0029] where P is a matrix composed of determination coefficients. When the actual flow rate is within the rolling forecast range, the determination coefficient is assigned 1; when it is outside the forecast range, the determination coefficient is assigned 0.

[0030] Preferably, in the process of index normalization in step (3): Since the value ranges of the various indexes obtained in step (2) are not the same, in order to obtain a comprehensive index in the next step, some indexes need to be normalized so that their value ranges are [0, 1].

[0031] Preferably, in step (4), the normalized indexes in step (3) are weighted and averaged, and the comprehensive evaluation index is obtained as follows:

[0032] Z = A·B

[0033] A = [a 1 a 2 a 3 a 4 a 5

[0034]

[0035] where A is the evaluation weight matrix, the sum of all elements in the matrix is 1, B is the normalized evaluation index matrix, and Z is the comprehensive evaluation index of the rolling forecast of the flood process this time, with a value range of 0 to 1.

[0036] Advantages of the present invention:

[0037] The present invention solves the problem that the existing single flood forecast process evaluation system is not applicable to the rolling forecast mode in actual production applications, makes up for the deficiencies of the single process evaluation system in evaluating the flood forecast effect in the rolling forecast mode, can comprehensively evaluate the effect of flood rolling forecast, and can better guide the flood forecast work in the actual production process. Description of the Drawings

[0038] Figure 1 It is a flood forecast process diagram under the rolling forecast mode of a certain station. Detailed Embodiment

[0039] The present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0040] Example 1: The flood forecast effect evaluation method based on the rolling forecast mode provided in this example includes the following steps:

[0041] Step (1): Data collection and collation. Collect and collate the actual flood process and the corresponding rolling forecast process data within a certain period to be evaluated.

[0042]

[0043] Among them, Q is the forecast flood process matrix, O is the measured flood process matrix, 1 to m are the flood process time periods, I to n are the number of rolling forecasts, and M and N are auxiliary matrices.

[0044] Step (2): Calculate 5 evaluation indicators. Include the mean process certainty coefficient (DC), average forecast dispersion (DI), mean relative error (MRE), peak relative error (MREP), and forecast coverage rate (CR).

[0045] ① Mean process certainty coefficient:

[0046] DC = NSE(Q', O)

[0047]

[0048] Among them, Q’ represents the rolling forecast average process corresponding to the actual flood process, NSE represents the calculation method of the certainty coefficient, reflecting the degree of coincidence between the prediction process Q' and the measured process O, and can refer to the "Hydrological Information and Forecast Specification (GB / T 22482-2008)".

[0049] ② Average forecast dispersion:

[0050]

[0051] Among them, Max represents taking the maximum value in the series, Min represents taking the minimum value in the series, and DI represents the average dispersion degree of each rolling forecast value in this flood process.

[0052] ③ Mean relative error:

[0053]

[0054] Among them, RE is the relative error matrix, and MRE is the mean relative error of the rolling forecast result.

[0055] ④ Peak mean error:

[0056]

[0057] Among them, O jis the peak value of the measured flood, j is its time series number, REP is the relative error matrix of the peak value, and MREP is the mean error of the peak value.

[0058] ⑤Forecast coverage:

[0059]

[0060] Among them, P is a matrix composed of determination coefficients. When the actual flow is within the rolling forecast range, the determination coefficient is assigned a value of 1, and when it is outside the forecast range, the determination coefficient is assigned a value of 0.

[0061] Step (3): Normalization of indicators. Since the value ranges of the indicators obtained in step (2) are not the same, in order to obtain a comprehensive evaluation indicator in the next step, some indicators need to be normalized so that their value range is [0, 1]. Preferably, the normalization method used by this method for each indicator is shown in Table 1.

[0062] Table 1 Normalization methods of various indicators

[0063]

[0064] Step (4): Calculate the comprehensive evaluation index. Perform a weighted average on the normalized indexes in step (3) to obtain the comprehensive evaluation index.

[0065] Z=A·B

[0066] A=[a 1 a 2 a 3 a 4 a 5 ]

[0067]

[0068] Among them, A is the evaluation weight matrix, and the sum of all elements in the matrix is ​​1. Preferably, all elements in this method are 0.2; B is the normalized evaluation index matrix; Z is the comprehensive evaluation index of the rolling forecast of this flood process, and its value range is [0,1]. The larger the value, the better the rolling forecast effect within this evaluation range.

[0069] Embodiment 2: A flood forecast effect evaluation method based on a rolling forecast mode, wherein the calculation method is as follows:

[0070] In this embodiment, if Figure 1 As shown in the figure, a certain hydrological station is taken as the research object. The hydrological station regularly issues flood forecasts at 8:00 every day. The forecast values ​​include the daily flow values ​​at 8:00 for the next 10 days, which is a typical rolling forecast mode.

[0071] An example of the evaluation method for flood forecasting effect based on the rolling forecasting mode is as follows:

[0072] (1) Data collection and collation.

[0073] Taking the flood process in August of a certain year at this hydrological station as the evaluation object, the forecast flood process matrix Q, the measured flood process matrix O, and the auxiliary matrices M and N are respectively obtained and sorted as follows.

[0074]

[0075]

[0076] (2) Calculate five evaluation indicators.

[0077] According to the relevant definitions and methods, five evaluation indicators are calculated respectively, including: the mean process certainty coefficient DC, the average forecast dispersion DI, the mean relative error MRE, the peak relative error MREP, and the forecast coverage rate CR.

[0078] DC = 0.865, DI = 0.209, MRE = 0.080, MREP = 0.135, CR = 0.903.

[0079] (3) Index normalization.

[0080] According to the normalization method shown in Table 1, the normalized values of each evaluation indicator are as follows:

[0081] DC’ = 0.865, DI’ = 0.791, MRE’ = 0.920, MREP’ = 0.865, CR’ = 0.903.

[0082] (4) Calculate the comprehensive evaluation indicator.

[0083] Set the weights of each evaluation indicator, and the comprehensive evaluation indicator value can be obtained after weighted averaging. In particular, in this embodiment, the weights of the evaluation indicators are all set to 0.2, and the comprehensive evaluation indicator value of the flood process based on the rolling forecasting mode in August at this hydrological station is 0.87. This value is relatively close to 1, indicating that the rolling forecasting effect in August at this hydrological station is good.

[0084] The above embodiments are only the preferred technical solutions of the present invention and should not be regarded as limitations on the present invention. The protection scope of the present invention should be the technical solutions recorded in the claims, including the equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, the equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A flood forecast effect evaluation method based on a rolling forecast model, characterized in that: It includes the following steps: Step (1): Data collection and collation: Collect and collate the actual flood process and the corresponding rolling forecast process data within a certain period of time to be evaluated; Step (2): Calculate five evaluation indicators: including mean process certainty coefficient, average forecast dispersion, average relative error, peak relative error and forecast coverage; Step (3): normalization of indicators; Step (4): Calculate the comprehensive evaluation index: Take the weighted average of the normalized indicators in step (3) to obtain the comprehensive evaluation index.

2. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: The data in step (1) includes the following: Among them, Q is the forecast flood process matrix, O is the measured flood process matrix, 1~m is the number of flood process periods, I~n is the number of rolling forecasts, and M and N are auxiliary matrices.

3. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: In step (2), the mean process certainty coefficient is: DC=NSE(Q',O) Among them, Q' represents the rolling forecast average process corresponding to the actual flood process, and NSE represents the calculation method of the certainty coefficient, which reflects the degree of agreement between the predicted process Q' and the measured process O.

4. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: In step (2), the average forecast dispersion is: Among them, Max means taking the maximum value in the series, Min means taking the minimum value in the series, and DI represents the average dispersion of each rolling forecast value during this flood process.

5. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: In step (2), the average relative error is: Among them, RE is the relative error matrix, and MRE is the mean relative error of the rolling forecast results.

6. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: In step (2), the peak average error is: Among them, O j is the peak value of the measured flood, j is its time series number, REP is the relative error matrix of the peak value, and MREP is the mean error of the peak value.

7. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: In step (2), the forecast coverage is: Among them, P is a matrix composed of determination coefficients. When the actual flow is within the rolling forecast range, the determination coefficient is assigned a value of 1, and when it is outside the forecast range, the determination coefficient is assigned a value of 0.

8. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: The specific process of normalizing the indicators in step (3) is as follows: Since the value ranges of the indicators obtained in step (2) are not the same, in order to obtain a comprehensive indicator in the next step, some indicators need to be normalized so that their value range is [0,1].

9. The flood forecast effect evaluation method based on the rolling forecast mode according to claim 1 is characterized in that: In step (4), the normalized indicators in step (3) are weighted averaged to obtain the comprehensive evaluation indicators as follows: Z=A·B A=[a1 a2 a3 a4 a5] Among them, A is the evaluation weight matrix, the sum of all elements in the matrix is ​​1, B is the normalized evaluation index matrix, and Z is the comprehensive evaluation index of the rolling forecast of this flood process, with a value of 0 to 1.

Images