poi3.8大数据量Excel导出连续查询导出不同的报表后导出速度非常慢

  poi3.8用((SXSSFSheet)sheet).flushRows()方法解决了大数据量的excel导出，但是我在使用过程中，由于每个单元格都要设置不同的颜色和数据格式，发现导出第一张报表时速度还勉强可以接受，但是继续导出另一张报表，速度就慢的跟蜗牛似的，要10分钟甚至3个小时，请问有什么方法可以解决？
        代码如下：

private static void fillExcel(KDTable kdtReport, SXSSFWorkbook book, Sheet sheet) {
int headRowCount = kdtReport.getHeadRowCount();
headRowCount=headRowCount+startRowCount;
Font font=book.createFont();
font.setFontHeightInPoints((short)10);
//改为保留整数
Short numFormat = book.createDataFormat().getFormat("0"); //numFormat==>0.00=0
//单元格样式或行样式
CellStyle cellStyle = book.createCellStyle(); // 建立新的cell样式
cellStyle.setBorderBottom(CellStyle.BORDER_THIN); //下边框

cellStyle.setBorderLeft(CellStyle.BORDER_THIN);//左边框

cellStyle.setBorderTop(CellStyle.BORDER_THIN);//上边框

cellStyle.setBorderRight(CellStyle.BORDER_THIN);//右边框

Set mergeBlockSet = new HashSet();

Cell cell = null;
Object value = null;
KDTMergeBlock mergeBlock = null;
IRow row2 = null;
Color color = null;
Row row = null;
int treeLevel = 0;
short colorIndex = IndexedColors.WHITE.getIndex();
for (int i = 0; i < kdtReport.getRowCount(); i++) {

row2 = kdtReport.getRow(i);
treeLevel = row2.getTreeLevel();
color = row2.getStyleAttributes().getBackground();
row = sheet.getRow(i + headRowCount);
if(row==null){
row = sheet.createRow(i+headRowCount);
}

cellStyle = book.createCellStyle();
cellStyle.setBorderBottom(CellStyle.BORDER_THIN); //下边框

cellStyle.setBorderLeft(CellStyle.BORDER_THIN);//左边框

cellStyle.setBorderTop(CellStyle.BORDER_THIN);//上边框

cellStyle.setBorderRight(CellStyle.BORDER_THIN);//右边框

cellStyle.setFillPattern(CellStyle.SOLID_FOREGROUND);

        if(Color.white.getRed()==color.getRed()&&Color.white.getGreen()==color.getGreen()&&Color.white.getBlue()==color.getBlue()){
            colorIndex = IndexedColors.WHITE.getIndex();
        }else{
            colorIndex = spaceColorIndex[treeLevel];
        }
        cellStyle.setFillPattern(CellStyle.SOLID_FOREGROUND);
        cellStyle.setFillForegroundColor(colorIndex);
        for (int j = 0; j < kdtReport.getColumnCount(); j++) {
            value = row2.getCell(j).getValue();                 
            mergeBlock = row2.getCell(j).getMergeBlock();
            if(mergeBlock!=null){
                mergeBlockSet.add(mergeBlock);
            }
            if (value != null) {
                if(KDTableUtil.TREE_LEVEL_COLKEY.equals(kdtReport.getColumn(j).getKey()) || "curProject".equals(kdtReport.getColumn(j).getKey())){
                    if(!RptHelper.IndexHistoryChangeUIName.equals(tableName)){//指标历史变化表(这个好特殊啊)
                        value = value.toString().trim();
                        value = space1[treeLevel]+value;
                    }
                }                   
                cell = row.createCell(j);
                if(value instanceof BigDecimal || value instanceof Integer || value instanceof Long || value instanceof Double){
                    BigDecimal bValue = FDCHelper.toBigDecimal(value);
                    if(bValue.compareTo(BigDecimal.ZERO)!=0){
                        cell.setCellValue(bValue.doubleValue());
                        cell.setCellType(XSSFCell.CELL_TYPE_NUMERIC);
                    }
                    cellStyle.setDataFormat(numFormat);                     
                    cell.setCellStyle(cellStyle);
                }else{
                    cell.setCellValue(value.toString());
                    cell.setCellType(XSSFCell.CELL_TYPE_STRING);    
                    if(cellStyle!=null){
                        cell.setCellStyle(cellStyle);
                    }else{
                        cellStyle = book.createCellStyle();

                        cell.setCellStyle(cellStyle);
                    }
                }
            }else{                  
                cell = row.createCell(j);   
                cellStyle.setFillForegroundColor(colorIndex);
                cell.setCellStyle(cellStyle);
            }   
        }               
        //每100行，就设置分组和合并信息，最后把这100行内容从内存写入本地硬盘
         if(i>1&&i%100==0){
             System.out.println("i:"+i);
             try {
                // 设置分组和合并
                setRowGroup(kdtReport, sheet);
                // 设置分组数据显示
                sheet.setRowSumsBelow(false);
                sheet.setRowSumsRight(false);
                //这100行内容从内存写入本地硬盘
                ((SXSSFSheet)sheet).flushRows();
                cellStyle = book.createCellStyle();
              } catch (IOException e) {                 
                e.printStackTrace();
              }

          }
    }
    //把剩余行内容从内存写入本地硬盘
    try {
        // 设置分组和合并
        setRowGroup(kdtReport, sheet);
        // 设置分组数据显示
        sheet.setRowSumsBelow(false);
        sheet.setRowSumsRight(false);
        ((SXSSFSheet)sheet).flushRows();
        cellStyle = null;
    } catch (IOException e) {           
        e.printStackTrace();
    }
    Iterator iterator = mergeBlockSet.iterator();
    CellRangeAddress cellRangeAddress =null;
    while (iterator.hasNext()) {
        mergeBlock = (KDTMergeBlock) iterator.next();

        int left = mergeBlock.getLeft();
        int right = mergeBlock.getRight();

        int top=mergeBlock.getTop()+headRowCount;
        int bottom=mergeBlock.getBottom()+headRowCount; 

        cellRangeAddress = new CellRangeAddress(top, bottom, left, right);
        sheet.addMergedRegion(cellRangeAddress);
    }
}