Short-term node electricity price prediction method based on node geographic position

Abstract

The invention relates to a short-term node electricity price prediction method based on a node geographic position, and belongs to the field of electricity price prediction. According to the method, multi-source data is collected, data preprocessing is carried out by using a quartile method, linear interpolation, Min-Max data normalization and other methods, space-time grid division is carried out based on node geographic position data, space-time grids are clustered by using a K-means clustering method, and a short-term node electricity price prediction model based on the improved LSTM is constructed. According to the method, space-time grid division is carried out based on the geographic position features of the nodes, power supply association mapping is established between each space grid and the nodes, the influence range of grid power utilization on the node electricity price is determined, a clustering sensing multi-branch LSTM prediction model is constructed, different loss functions are adopted in different grids, and the prediction precision of the node electricity price is effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of electricity price forecasting, specifically relating to a short-term nodal electricity price forecasting method based on the geographical location of nodes. Background Technology

[0002] With the continuous development of the electricity market, electricity prices, as a core element, are crucial for the accurate prediction of market participants' decisions, the economic operation of the power system, and the rational allocation of energy resources. Short-term electricity prices are influenced by various factors, such as historical price data, power load, weather conditions, holidays, policy adjustments, and fuel prices, exhibiting strong volatility, nonlinearity, and uncertainty. At different nodes in the power system, mismatches between available transmission capacity and renewable energy unit output can lead to node congestion and drastic fluctuations in node prices. Existing short-term electricity price prediction methods often have significant errors in these special scenarios, making them unsuitable for practical applications. Therefore, a method to improve the accuracy of short-term electricity price prediction in special scenarios is urgently needed. Summary of the Invention

[0003] The purpose of this invention is to provide a short-term nodal electricity price forecasting method based on the geographical location of nodes, so as to solve the problem that existing electricity price forecasting schemes have large errors at different nodes.

[0004] To achieve the above objectives, the technical solution of the present invention is: a short-term nodal electricity price prediction method based on node geographical location, comprising the following steps:

[0005] Step S1: Multi-source data acquisition and preprocessing, including node geographical location data, historical node electricity price data, electricity load data of the power supply area corresponding to the node, distributed power output data, historical weather data, holiday schedule, and regional economic activity data;

[0006] Step S2: Data preprocessing, using the quartile method to remove outliers, using linear interpolation to complete missing data, and using Min-Max data normalization;

[0007] Step S3: Spatiotemporal grid partitioning based on node geographic location data;

[0008] Step S4: K-means clustering of spatiotemporal grid features;

[0009] Step S5: Construct a short-term nodal electricity price prediction model based on an improved LSTM.

[0010] Furthermore, step S3 is implemented as follows:

[0011] Step S31, Geographic Coordinate Mapping: Using the Mercator projection method, map the geographic coordinates of the node power supply area to a plane coordinate system;

[0012] Step S32, Spatial Grid Division: Based on the mapped planar coordinates, load fluctuation rate is used and a threshold is set to divide the node power supply area into two types of spatial grids: dedicated grid: corresponding to areas with high load fluctuation rate, including commercial complexes and industrial parks with high electricity load and concentrated population flow; comprehensive grid: corresponding to areas with low load fluctuation rate, including residential areas and ordinary blocks; set the edge length benchmark value to calculate the grid edge length.

[0013] Step S33, Time Dimension Division: Based on the regional electricity load fluctuation characteristics, the day is divided into multiple time slices, each time slice corresponds to a time dimension unit, which is combined with the spatial grid to form a two-dimensional spatiotemporal grid.

[0014] Furthermore, in step S32, the method for calculating the load fluctuation rate is as follows:

[0015]

[0016] in, The load fluctuation rate within the node's power supply area. , , These are the daily maximum load, daily minimum load, and daily average load, respectively.

[0017] The method for calculating the grid side length is as follows:

[0018]

[0019] in, , , , These are the grid side length, grid side length baseline value, grid side length correction factor, and grid load fluctuation rate for the dedicated grid area;

[0020]

[0021] in, , , , These are the grid side length, grid side length baseline value, grid side length correction coefficient, and grid load fluctuation rate of the integrated grid region, respectively.

[0022] Furthermore, in step S31, the calculation formula for the Mercator projection method is as follows:

[0023]

[0024] in, The x-coordinate of the geographic entity node after projection, For the radius of the Earth, Longitude of nodal geographic entities The longitude of the Earth's central meridian;

[0025]

[0026] in, The vertical coordinates of the projected node geographic entities. The latitude of the node geographic entity.

[0027] Furthermore, in step S4, for each spatiotemporal grid, the feature vector coupling electricity consumption status and electricity price within its region is extracted, and K-means clustering is performed. The specific implementation is as follows:

[0028] Step S41: Randomly select n objects from the data collected in step S1, and use them as the cluster centers of the n classes respectively;

[0029] Step S42: Assign other data to the class closest to it, resulting in n cluster groups; use Euclidean distance to measure the similarity between a sample and its cluster center, calculated using the following formula:

[0030]

[0031] in, , f and f are the f-th dimension of the nth grid feature vector in the s-th season and the f-th dimension of the nth cluster center vector in the s-th season, respectively. Let be the distance between the nth grid feature vector and the nth cluster center vector in the s-th season. Let be the weight of the s-th season;

[0032] Step S43: Recalculate the centroid of each cluster;

[0033] Step S44: Continue to divide and classify based on the new cluster centroids, repeating steps S42 and S43 until the sum of squared errors of the centroids is minimized. The specific formula is as follows:

[0034]

[0035] in, The number of samples per cluster;

[0036] Step S45: Use silhouette coefficient to evaluate the clustering effect, as follows:

[0037]

[0038] Let be the silhouette coefficient of the i-th sample. Let be the average distance from the i-th sample to the other samples in the same cluster. For the i-th sample to other clusters C j The average distance between all samples is called the distance between sample i and cluster C. j Dissimilarity:

[0039] The spatiotemporal grids are divided into three categories by clustering: high electricity price and high load, medium electricity price and medium load, and low electricity price and low load. Each category of spatiotemporal grids corresponds to a node electricity price influence pattern.

[0040] Furthermore, in step S5, the short-term nodal electricity price prediction model based on the improved LSTM includes a category-aware attention module and a multi-branch LSTM network; each branch of the LSTM network includes an input gate, an output gate, and a forget gate.

[0041] The category-aware attention module uses analogy-guided Softmax attention, as shown below:

[0042]

[0043] in, The f-th dimension of the original feature of the m-th cluster category is... Let m be the embedding vector of the m-th cluster category. The attention weights are the feature vectors of the f-th dimension of the m-th cluster category. It is a Layer 1 fully connected network;

[0044] The final input features for the m-th cluster category are:

[0045]

[0046] Let f be the final input feature of the m-th cluster category;

[0047] The multi-branch LSTM network includes high-price, high-load branches, medium-price, medium-load branches, and low-price, low-load branches. Different branches use different loss functions, as detailed below:

[0048] Weighted mean square error is used for branches with high electricity prices and high loads:

[0049]

[0050] in, To determine the sample size for high-electricity-price, high-load areas, For the actual electricity price of i samples in a high-electricity-price, high-load area, The predicted electricity price for i samples in a high-electricity-price, high-load region; This is a statistical indicator of the error between predicted and actual electricity prices in samples from high-electricity-price, high-load areas. The weight of the prediction error for the i-th sample;

[0051] The root mean square error is used for load branches in the medium electricity price:

[0052]

[0053] This represents the sample size for the load area within the medium electricity price range. Let j be the actual electricity prices of the load areas in the medium-priced electricity market. Let j be the predicted electricity prices for the load areas in the medium-priced electricity market;

[0054] Root mean square error used in low-electricity-price, low-load branches:

[0055]

[0056] The sample size for low-electricity-price, low-load areas. For j samples in a low-electricity-price, low-load area, the actual electricity price is... The predicted electricity price for j samples in a low-electricity-price, low-load region;

[0057] The input gates in each branch of the LSTM network, according to Key information was extracted from load fluctuations and historical electricity price data, as shown below:

[0058]

[0059] in, For each branch of the LSTM network, the activation function is... The weights of the input gates, The electrical load of the power supply area at time t-1 is the value of the load. The quantity of electrical load in the power supply area of ​​node t-1. This is the bias of the input gate. For the cell states in each branch of the LSTM network, is the activation function for the input gate;

[0060] Forget gates in each branch of the LSTM network, discarding... The short-term noise in the data is shown below:

[0061]

[0062] in, The weight of the Forgotten Gate The bias of the forgetting gate;

[0063] The output gates in each branch of the LSTM network use fully connected layers to output the predicted node electricity price of the corresponding branch at the current time based on the updated cell state.

[0064] The present invention also provides a short-term nodal electricity price forecasting system based on node geographical location, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it can implement the steps of the method described above.

[0065] The present invention also provides an electronic device including a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method described above.

[0066] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0067] Compared with existing technologies, this invention has the following advantages: This invention uses spatiotemporal grid partitioning based on node geographic location features to transform high-dimensional, heterogeneous spatiotemporal data into low-dimensional, structured, and strongly correlated analysis units through spatial functionalization and temporal regularization. This provides accurate and efficient input for subsequent clustering and prediction models, ultimately improving the feature capture capability and prediction accuracy of target variables in short-term prediction scenarios. Based on node geographic location features, spatiotemporal grid partitioning establishes a power supply association mapping between each spatial grid and the node, clarifying the impact range of grid electricity consumption on node electricity prices. A cluster-aware multi-branch LSTM prediction model is constructed, employing different loss functions in different grids, effectively improving the prediction accuracy of node electricity prices. It possesses the characteristics of multi-feature fusion, dynamic association capture, and strong noise robustness, providing stable and accurate time-series prediction results in scenarios with strong spatiotemporal dynamics of short-term node electricity prices and high prediction accuracy requirements. Attached Figure Description

[0068] Figure 1 This is a flowchart of the short-term nodal electricity price prediction process based on the geographical location of the node according to the present invention.

[0069] Figure 2 This is a flowchart of the K-means clustering process of the present invention. Detailed Implementation

[0070] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0071] like Figure 1 , 2As shown, this embodiment of the invention provides a short-term nodal electricity price prediction method based on node geographical location, including the following steps:

[0072] Step S1: Multi-source data acquisition and preprocessing, including node geographical location data, historical node electricity price data, electricity load data of the power supply area corresponding to the node, distributed power output data, historical weather data, holiday schedule, and regional economic activity data;

[0073] Step S2: Data preprocessing, using the quartile method to remove outliers, using linear interpolation to complete missing data, and using Min-Max data normalization;

[0074] Step S3: Spatiotemporal grid partitioning based on node geographic location data; the specific implementation is as follows:

[0075] Step S31, Geographic Coordinate Mapping: Using the Mercator projection method, map the geographic coordinates of the node power supply area to a plane coordinate system;

[0076] The calculation formula for the Mercator projection method is as follows:

[0077]

[0078] in, The x-coordinate of the geographic entity node after projection, For the radius of the Earth, Longitude of nodal geographic entities The longitude of the Earth's central meridian;

[0079]

[0080] in, The vertical coordinates of the projected node geographic entities. The latitude of the node geographic entity;

[0081] Step S32, Spatial Grid Division: Based on the mapped planar coordinates, load fluctuation rate is used and a threshold is set to divide the node power supply area into two types of spatial grids: dedicated grid: corresponding to areas with high load fluctuation rate, including commercial complexes and industrial parks with high electricity load and concentrated population flow; comprehensive grid: corresponding to areas with low load fluctuation rate, including residential areas and ordinary blocks; set the edge length benchmark value to calculate the grid edge length.

[0082] The method for calculating load volatility is as follows:

[0083]

[0084] in, The load fluctuation rate within the node's power supply area. , , These are the daily maximum load, daily minimum load, and daily average load, respectively.

[0085] The method for calculating the grid side length is as follows:

[0086]

[0087] in, , , , These are the grid side length, grid side length baseline value, grid side length correction factor, and grid load fluctuation rate for the dedicated grid area;

[0088]

[0089] in, , , , These are the grid side length, grid side length baseline value, grid side length correction factor, and grid load fluctuation rate of the integrated grid region, respectively.

[0090] Step S33, Time Dimension Division: Based on the regional electricity load fluctuation characteristics, the day is divided into multiple time slices, each time slice corresponds to a time dimension unit, which is combined with the spatial grid to form a two-dimensional spatiotemporal grid.

[0091] Step S4: K-means clustering of spatiotemporal grid features; the specific implementation is as follows:

[0092] For each spatiotemporal grid, the feature vector coupling electricity consumption status and electricity price within its region is extracted, and K-means clustering is performed. The specific implementation is as follows:

[0093] Step S41: Randomly select n objects from the data collected in step S1, and use them as the cluster centers of the n classes respectively;

[0094] Step S42: Assign other data to the class closest to it, resulting in n cluster groups; use Euclidean distance to measure the similarity between a sample and its cluster center, calculated using the following formula:

[0095]

[0096] in, , f and f are the f-th dimension of the nth grid feature vector in the s-th season and the f-th dimension of the nth cluster center vector in the s-th season, respectively. Let be the distance between the nth grid feature vector and the nth cluster center vector in the s-th season. Let be the weight of the s-th season;

[0097] Step S43: Recalculate the centroid of each cluster;

[0098] Step S44: Continue to divide and classify based on the new cluster centroids, repeating steps S42 and S43 until the sum of squared errors of the centroids is minimized. The specific formula is as follows:

[0099]

[0100] in, The number of samples per cluster;

[0101] Step S45: Use silhouette coefficient to evaluate the clustering effect, as follows:

[0102]

[0103] Let be the silhouette coefficient of the i-th sample. Let be the average distance from the i-th sample to the other samples in the same cluster. For the i-th sample to other clusters C j The average distance between all samples is called the distance between sample i and cluster C. j Dissimilarity:

[0104] The spatiotemporal grids are divided into three categories by clustering: high electricity price and high load, medium electricity price and medium load, and low electricity price and low load. Each category of spatiotemporal grids corresponds to a node electricity price influence pattern.

[0105] Step S5: Constructing a short-term nodal electricity price prediction model based on an improved LSTM; the specific implementation is as follows:

[0106] The short-term nodal electricity price prediction model based on the improved LSTM includes a category-aware attention module and a multi-branch LSTM network; each branch of the LSTM network includes an input gate, an output gate, and a forget gate.

[0107] The category-aware attention module uses analogy-guided Softmax attention, as shown below:

[0108]

[0109] in, The f-th dimension of the original feature of the m-th cluster category is... Let m be the embedding vector of the m-th cluster category. The attention weights are the feature vectors of the f-th dimension of the m-th cluster category. It is a Layer 1 fully connected network;

[0110] The final input features for the m-th cluster category are:

[0111]

[0112] Let f be the final input feature of the m-th cluster category;

[0113] The multi-branch LSTM network includes high-price, high-load branches, medium-price, medium-load branches, and low-price, low-load branches. Different branches use different loss functions, as detailed below:

[0114] Weighted mean square error is used for branches with high electricity prices and high loads:

[0115]

[0116] in, To determine the sample size for high-electricity-price, high-load areas, For the actual electricity price of i samples in a high-electricity-price, high-load area, The predicted electricity price for i samples in a high-electricity-price, high-load region; This is a statistical indicator of the error between predicted and actual electricity prices in samples from high-electricity-price, high-load areas. The weight of the prediction error for the i-th sample;

[0117] The root mean square error is used for load branches in the medium electricity price:

[0118]

[0119] This represents the sample size for the load area within the medium electricity price range. Let j be the actual electricity prices of the load areas in the medium-priced electricity market. Let j be the predicted electricity prices for the load areas in the medium-priced electricity market;

[0120] Root mean square error used in low-electricity-price, low-load branches:

[0121]

[0122] The sample size for low-electricity-price, low-load areas. For j samples in a low-electricity-price, low-load area, the actual electricity price is... The predicted electricity price for j samples in a low-electricity-price, low-load region;

[0123] The input gates in each branch of the LSTM network, according to Key information was extracted from load fluctuations and historical electricity price data, as shown below:

[0124]

[0125] in, For each branch of the LSTM network, the activation function is... The weights of the input gates, The electrical load of the power supply area at time t-1 is the value of the load. The quantity of electrical load in the power supply area of ​​node t-1. This is the bias of the input gate. For the cell states in each branch of the LSTM network, is the activation function for the input gate;

[0126] Forget gates in each branch of the LSTM network, discarding... The short-term noise in the data is shown below:

[0127]

[0128] in, The weight of the Forgotten Gate The bias of the forgetting gate;

[0129] The output gates in each branch of the LSTM network use fully connected layers to output the predicted node electricity price of the corresponding branch at the current time based on the updated cell state.

[0130] The present invention also provides a short-term nodal electricity price forecasting system based on node geographical location, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it can implement the steps of the method described above.

[0131] The present invention also provides an electronic device including a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method described above.

[0132] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0133] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for short-term nodal electricity price forecasting based on node geographic location, characterized in that, Includes the following steps: Step S1: Multi-source data acquisition and preprocessing, including node geographical location data, historical node electricity price data, electricity load data of the power supply area corresponding to the node, distributed power output data, historical weather data, holiday schedule, and regional economic activity data; Step S2: Data preprocessing, using the quartile method to remove outliers, using linear interpolation to complete missing data, and using Min-Max data normalization; Step S3: Spatiotemporal grid partitioning based on node geographic location data; Step S4: K-means clustering of spatiotemporal grid features; Step S5: Construct a short-term nodal electricity price prediction model based on an improved LSTM.

2. The method for short-term nodal electricity price forecasting based on nodal geographical location according to claim 1, characterized in that, Step S3 is implemented as follows: Step S31, Geographic Coordinate Mapping: Using the Mercator projection method, map the geographic coordinates of the node power supply area to a plane coordinate system; Step S32, Spatial Grid Division: Based on the mapped planar coordinates, load fluctuation rate is used and a threshold is set to divide the node power supply area into two types of spatial grids: dedicated grid: corresponding to areas with high load fluctuation rate, including commercial complexes and industrial parks with high electricity load and concentrated population flow; comprehensive grid: corresponding to areas with low load fluctuation rate, including residential communities and ordinary blocks. Set the baseline value for the side length to calculate the grid side length; Step S33, Time Dimension Division: Based on the regional electricity load fluctuation characteristics, the day is divided into multiple time slices, each time slice corresponds to a time dimension unit, which is combined with the spatial grid to form a two-dimensional spatiotemporal grid.

3. The method for short-term nodal electricity price forecasting based on nodal geographical location according to claim 2, characterized in that, In step S32, the method for calculating the load fluctuation rate is as follows: in, The load fluctuation rate within the node's power supply area. , , These are the daily maximum load, daily minimum load, and daily average load, respectively. The method for calculating the grid side length is as follows: in, , , , These are the grid side length, grid side length baseline value, grid side length correction factor, and grid load fluctuation rate for the dedicated grid area; in, , , , These are the grid side length, grid side length baseline value, grid side length correction coefficient, and grid load fluctuation rate of the integrated grid region, respectively.

4. The method for short-term nodal electricity price forecasting based on nodal geographical location according to claim 2, characterized in that, In step S31, the calculation formula for the Mercator projection method is as follows: in, The x-coordinate of the geographic entity node after projection, For the radius of the Earth, Longitude of nodal geographic entities The longitude of the Earth's central meridian; in, The vertical coordinates of the projected node geographic entities. The latitude of the node geographic entity.

5. The method for short-term nodal electricity price forecasting based on nodal geographical location according to claim 1, characterized in that, In step S4, for each spatiotemporal grid, the feature vector coupling electricity consumption status and electricity price within its region is extracted, and K-means clustering is performed. The specific implementation is as follows: Step S41: Randomly select n objects from the data collected in step S1, and use them as the cluster centers of the n classes respectively; Step S42: Assign other data to the class closest to it, resulting in n cluster groups; use Euclidean distance to measure the similarity between a sample and its cluster center, calculated using the following formula: in, , f and f are the f-th dimension of the nth grid feature vector in the s-th season and the f-th dimension of the nth cluster center vector in the s-th season, respectively. Let be the distance between the nth grid feature vector and the nth cluster center vector in the s-th season. Let be the weight of the s-th season; Step S43: Recalculate the centroid of each cluster; Step S44: Continue to divide and classify based on the new cluster centroids, repeating steps S42 and S43 until the sum of squared errors of the centroids is minimized. The specific formula is as follows: in, The number of samples per cluster; Step S45: Use silhouette coefficient to evaluate the clustering effect, as follows: Let be the silhouette coefficient of the i-th sample. Let be the average distance from the i-th sample to the other samples in the same cluster. For the i-th sample to other clusters C j The average distance between all samples is called the distance between sample i and cluster C. j Dissimilarity: The spatiotemporal grids are divided into three categories by clustering: high electricity price and high load, medium electricity price and medium load, and low electricity price and low load. Each category of spatiotemporal grids corresponds to a node electricity price influence pattern.

6. The method for short-term nodal electricity price forecasting based on nodal geographical location according to claim 1, characterized in that, In step S5, the short-term nodal electricity price prediction model based on the improved LSTM includes a category-aware attention module and a multi-branch LSTM network; each branch of the LSTM network includes an input gate, an output gate, and a forget gate.

7. A method for short-term nodal electricity price forecasting based on nodal geographical location according to claim 1 or 6, characterized in that, Step S5 is implemented as follows: The category-aware attention module uses analogy-guided Softmax attention, as shown below: in, The f-th dimension of the original feature of the m-th cluster category is... Let m be the embedding vector of the m-th cluster category. The attention weights are the feature vectors of the f-th dimension of the m-th cluster category. It is a Layer 1 fully connected network; The final input features for the m-th cluster category are: Let f be the final input feature of the m-th cluster category; The multi-branch LSTM network includes high-price, high-load branches, medium-price, medium-load branches, and low-price, low-load branches. Different branches use different loss functions, as detailed below: Weighted mean square error is used for branches with high electricity prices and high loads: in, To determine the sample size for high-electricity-price, high-load areas, For the actual electricity price of i samples in a high-electricity-price, high-load area, The predicted electricity price for i samples in a high-electricity-price, high-load region; This is a statistical indicator of the error between predicted and actual electricity prices in samples from high-electricity-price, high-load areas. The weight of the prediction error for the i-th sample; The root mean square error is used for load branches in the medium electricity price: This represents the sample size for the load area within the medium electricity price range. Let j be the actual electricity prices of the load areas in the medium-priced electricity market. Let j be the predicted electricity prices for the load areas in the medium-priced electricity market; Root mean square error used in low-electricity-price, low-load branches: The sample size for low-electricity-price, low-load areas. For j samples in a low-electricity-price, low-load area, the actual electricity price is... The predicted electricity price for j samples in a low-electricity-price, low-load region; The input gates in each branch of the LSTM network, according to Key information was extracted from load fluctuations and historical electricity price data, as shown below: in, For each branch of the LSTM network, the activation function is... The weights of the input gates, The electrical load of the power supply area at time t-1 is the value of the load. The quantity of electrical load in the power supply area of ​​node t-1. This is the bias of the input gate. For the cell states in each branch of the LSTM network, is the activation function for the input gate; Forget gates in each branch of the LSTM network, discarding... The short-term noise in the data is shown below: in, The weight of the Forgotten Gate The bias of the forgetting gate; The output gates in each branch of the LSTM network use fully connected layers to output the predicted node electricity price of the corresponding branch at the current time based on the updated cell state.

8. A short-term nodal electricity price forecasting system based on node geographic location, characterized in that, It includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor, which, when executed by the processor, enable the implementation of the steps of the method as described in any one of claims 1-7.

9. An electronic device comprising a processor and a memory, wherein, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium having stored thereon computer program instructions executable by a processor, wherein when the processor executes the computer program instructions, it is able to implement the steps of the method as described in any one of claims 1-7.

Images